Executing batch jobs in a distributed cloud environment using machine learning

Abstract

The method includes estimating the number of threads that can run a batch job within the maximum completion time by running a thread estimation algorithm. Running the thread estimation algorithm includes using an ML model to determine one or more combinations of threads, each containing the estimated number of threads and the actual completion time for processing the batch job using the estimated number of threads, wherein the actual completion time is less than or equal to the maximum completion time, and further includes storing the set containing one or more combinations in a thread and completion time list. The thread and completion time list is used to select the optimal virtual machine or optimal virtual machine combination based on the list of available cloud virtual machines, the optimal virtual machine or optimal virtual machine combination being able to run the batch job within the maximum completion time.

Description

Technical Field

[0004] , , , , ,

[0003] ,

[0001] Field The present disclosure generally relates to artificial intelligence technology, and more particularly to estimating cloud resources for batch processing using machine learning.

Background Art

[0002] Background Machine learning (ML) is an area of artificial intelligence (AI) where a computer can learn without being explicitly programmed. There are various types of ML techniques, such as supervised learning techniques and unsupervised learning techniques. In supervised learning techniques, an ML model is trained using training data, which includes a plurality of training examples, and each of the plurality of training examples includes an input and a known output corresponding to the input. In unsupervised learning techniques, unlabeled data is provided to an ML model or algorithm, and the purpose is to analyze the unlabeled data to find patterns. Examples of unsupervised learning techniques are dimensionality reduction and clustering.

[0003] As part of training, the model to be trained learns the function of mapping the inputs in the training data to their respective corresponding known outputs. When the model is appropriately trained by the training data, it can be used for output prediction for new inputs whose outputs are unknown. This is often referred to as the inference phase.

[0004] Models trained using supervised learning techniques are typically used to solve two types of tasks: classification tasks and regression tasks. In classification tasks, as part of training, the model learns to map inputs to output values, which derive from a discrete set of classes (or class labels). Thus, in classification problems, the model learns to map inputs to one or more classes within a set of discrete classes. In regression tasks, the model learns to map inputs to continuous output values ​​(e.g., continuous real numbers).

[0005] In computing, batch processing refers to the execution of a group of programs, or jobs, that do not involve user interaction. For example, many businesses use batch processing to perform relatively large amounts of computer work.

[0006] Furthermore, some businesses, due to the nature of their tasks, cannot execute and / or complete parts of them in real time. For example, in the banking industry, there are tasks that can only be completed at the end of a specific day, such as quarterly interest calculations. Therefore, tasks that need to be completed during a certain time period, for example, after working hours, are executed by combining multiple jobs, for example, through batch processing.

[0007] Therefore, for some businesses, batch processing is a time-dependent process that must be completed within a specific time frame by a deadline, for example, after business hours but before midnight. Due to the time dependency of the process, such businesses try to ensure that sufficient computing resources are available during certain time periods. [Overview of the project] [Problems that the invention aims to solve]

[0008] However, it is difficult to meet batch processing requirements with sufficient available resources without over-allocating resources. In some cases, the computer network becomes overloaded, and computing resources are unnecessarily wasted. [Means for solving the problem]

[0009] overview The technologies disclosed herein generally relate to the use of artificial intelligence technologies. More specifically and non-limitingly, the technologies disclosed herein relate to novel methods for using machine learning to estimate cloud resources for executing batch jobs. Various embodiments are described herein to illustrate various features. These embodiments include various methods, systems, non-temporary computer-readable storage media for storing programs, code, or instructions executable by one or more processors.

[0010] In various embodiments, a computer execution method is provided which is performed by a computer system. The computer execution method includes estimating the number of threads to execute a batch job within a maximum completion time by executing a thread estimation algorithm, which includes determining a set of one or more combinations using a machine learning (ML) model, each of which includes the estimated number of threads and the actual completion time for processing the total number of records of the batch job using the estimated number of threads, the actual completion time being less than or equal to the maximum completion time, and further includes storing the set of one or more combinations in a threads and completion time list, which is used to select an optimal virtual machine or optimal virtual machine combination based on an available cloud virtual machine list, the optimal virtual machine or optimal virtual machine combination being able to execute the batch job within the maximum completion time.

[0011] In some embodiments, executing a thread estimation algorithm further includes: determining the thread initialization time for each of a plurality of threads from 2 to a predetermined maximum number of threads, based on the current number of threads to process the batch job in the current iteration of a plurality of iterations; determining the process execution time using the maximum completion time and the thread initialization time; providing the process execution time as input to an ML model; obtaining the predicted number of records to be processed by the current number of threads as output to the ML model based on the input process execution time; determining whether the predicted number of records is greater than or equal to the expected number of records that can be processed using the current number of threads; if the predicted number of records is greater than or equal to the expected number of records, using the ML model to determine the actual completion time to process the expected number of records, and storing the current number of threads corresponding to the estimated number of threads in the current iteration in association with the actual completion time of the batch job in a thread and completion time list; and outputting a thread and completion time list based on a plurality of iterations performed for a plurality of threads.

[0012] In some embodiments, determining the actual completion time for processing an expected number of records using an ML model includes providing the expected number of records to be processed in the current iteration as input to the ML model, where the expected number of records is calculated by dividing the total number of records by the current number of threads, obtaining the execution time corresponding to the expected number of records as output of the ML model, and further calculating the actual completion time by adding the thread initialization time to the execution time output by the ML model.

[0013] In some embodiments, the computer execution method further includes selecting a number of maximum efficiency single virtual machines from a cloud virtual machine list so that each of the multiple configurations includes a number of maximum efficiency single virtual machines available in each of the multiple configurations, and the computer execution method further includes placing the number of maximum efficiency single virtual machines into a list of multiple maximum efficiency single virtual machines.

[0014] In some embodiments, one or more combinations in the thread and completion time list include one combination, the computer execution method includes determining a set of virtual machines comprising a plurality of high-efficiency virtual resources, each of the plurality of high-efficiency virtual resources includes one of a single virtual machine and a combination of virtual machines determined from a plurality of maximum-efficiency single virtual machines, each of the plurality of high-efficiency virtual resources has a number of cores greater than or equal to the total number of cores that can run a batch job within the maximum completion time, and the computer execution method further includes storing the plurality of high-efficiency virtual resources in a high-efficiency virtual resource list in descending order of efficiency, and selecting the optimal virtual machine or optimal combination of virtual machines from the high-efficiency virtual resource list as the maximum-efficiency virtual resource.

[0015] In some embodiments, selecting the optimal virtual machine or optimal combination of virtual machines as the most efficient virtual resource includes determining the lowest-cost virtual machine or combination of virtual machines from a list of highly efficient virtual resources.

[0016] In some embodiments, one or more combinations in the thread and completion time list include a plurality of combinations, and the computer execution method further includes iteratively determining for each of the plurality of combinations a set of virtual machines including a plurality of highly efficient virtual resources, each of the plurality of highly efficient virtual resources includes one of a single virtual machine and a virtual machine combination determined from a plurality of maximum efficient single virtual machines, each of the plurality of highly efficient virtual resources has a number of cores greater than or equal to the total number of cores that can run a batch job within the maximum completion time, and the computer execution method further includes iteratively storing the plurality of highly efficient virtual resources in a highly efficient virtual resource list in descending order of efficiency, and selecting the maximum efficient virtual resource as the optimal virtual machine or optimal virtual machine combination based on the plurality of highly efficient virtual resource lists obtained based on the execution of a plurality of iterations.

[0017] In some embodiments, the computer execution method further includes generating a training dataset using training data associated with a certain type of historical batch job, and generating an ML model for processing a certain type of batch job by training an input model using the training dataset.

[0018] In some embodiments, the ML model includes a regression model. In various embodiments, a computer execution method is provided which is performed by a computer system. The computer execution method includes obtaining a list of threads and completion times for processing a batch job within a maximum completion time, the list of threads and completion times includes a set of one or more combinations, each of which includes a thread estimate and an actual completion time for processing the total number of records of the batch job using the thread estimate, and the computer execution method further includes using the list of threads and completion times to select an optimal virtual machine or optimal virtual machine combination from a list of available cloud virtual machines, the optimal virtual machine or optimal virtual machine combination being capable of executing the batch job within a maximum completion time.

[0019] In some embodiments, selecting an optimal virtual machine or optimal virtual machine combination further includes selecting multiple maximum efficiency single virtual machines from a cloud virtual machine list so that each of the multiple configurations includes a maximum efficiency single virtual machine available in each of the multiple configurations, and selecting an optimal virtual machine or optimal virtual machine combination further includes placing the multiple maximum efficiency single virtual machines in a list of multiple maximum efficiency single virtual machines.

[0020] In some embodiments, one or more combinations in the thread and completion time list include one combination, selecting an optimal virtual machine or optimal virtual machine combination includes determining a set of virtual machines including multiple high-efficiency virtual resources, each of the multiple high-efficiency virtual resources includes one of the single virtual machines and virtual machine combinations determined from a list of multiple maximum-efficiency single virtual machines, each of the multiple high-efficiency virtual resources has a number of cores greater than or equal to the total number of cores required to run batch jobs within the maximum completion time, and selecting an optimal virtual machine or optimal virtual machine combination further includes storing the multiple high-efficiency virtual resources in a high-efficiency virtual resource list in descending order of efficiency, and selecting an optimal virtual machine or optimal virtual machine combination from the high-efficiency virtual resource list as the maximum-efficiency virtual resource.

[0021] In some embodiments, selecting the optimal virtual machine or optimal combination of virtual machines as the most efficient virtual resource includes determining the lowest-cost virtual machine or combination of virtual machines from a list of highly efficient virtual resources.

[0022] In some embodiments, one or more combinations in the thread and completion time list include multiple combinations, and selecting the optimal virtual machine or optimal virtual machine combination further includes iteratively determining for each of the multiple combinations a set of virtual machines including multiple high-efficiency virtual resources, each of the multiple high-efficiency virtual resources including one of the single virtual machines and virtual machine combinations determined from a list of multiple maximum-efficiency single virtual machines, each of the multiple high-efficiency virtual resources having a number of cores greater than or equal to the total number of cores that can run a batch job within the maximum completion time, and selecting the optimal virtual machine or optimal virtual machine combination further includes iteratively saving the multiple high-efficiency virtual resources in a high-efficiency virtual resource list in descending order of efficiency, and selecting the maximum-efficiency virtual resource as the optimal virtual machine or optimal virtual machine combination based on the multiple high-efficiency virtual resource lists obtained based on the execution of multiple iterations.

[0023] In some embodiments, obtaining a list of threads and completion times further includes, for each of a number of threads from 2 to a predetermined maximum number of threads, determining the thread initialization time based on the current number of threads to process the batch job in the current iteration of a plurality of iterations; determining the process execution time using the maximum completion time and the thread initialization time; providing the process execution time as input to a machine learning (ML) model; obtaining the predicted number of records to be processed by the current number of threads as output of the ML model based on the input process execution time; determining whether the predicted number of records is greater than or equal to the expected number of records that can be processed using the current number of threads; if the predicted number of records is greater than or equal to the expected number of records, using the ML model to determine the actual completion time to process the expected number of records, and storing the current number of threads corresponding to the estimated number of threads in the current iteration in association with the actual completion time of the batch job in a list of threads and completion times; and outputting a list of threads and completion times based on a plurality of iterations performed for a plurality of threads.

[0024] In some embodiments, determining the actual completion time for processing an expected number of records using an ML model includes providing the expected number of records processed in the current iteration and calculated by dividing the total number of records by the current number of threads as input to the ML model, obtaining the execution time corresponding to the expected number of records as output of the ML model, and calculating the actual completion time by adding the thread initialization time to the execution time output by the ML model.

[0025] In some embodiments, the computer execution method further includes generating a training dataset using training data associated with a certain type of historical batch job, and generating an ML model for processing a certain type of batch job by training an input model using the training dataset.

[0026] In some embodiments, the ML model includes a regression model. Some embodiments of the present disclosure include a system comprising one or more data processors. In some embodiments, when executed on one or more data processors, the system comprises a non-transitory computer-readable storage medium including instructions that cause the one or more data processors to execute some or all of one or more of the methods and / or some or all of one or more of the processes disclosed herein.

[0027] Some embodiments of the present disclosure include a computer program product embodied in a non-transitory machine-readable storage medium and including instructions configured to cause one or more data processors to execute some or all of one or more of the methods and / or some or all of one or more of the processes disclosed herein.

[0028] The above and following techniques can be realized in many ways and in many contexts. As will be described in more detail below, several exemplary embodiments and contexts are provided with reference to the following drawings. However, the following embodiments and contexts are only a part of many.

Brief Description of the Drawings

[0029] [Figure 1A] It is a simplified block diagram of a resource estimation system according to various embodiments. [Figure 1B] It is a simplified block diagram of a resource estimation system in a distributed computing environment according to various embodiments. [Figure 1C] It is a simplified block diagram of a resource estimation system in a cloud service provider (CSP) infrastructure according to various embodiments. [Figure 2A] It is a diagram showing an example of a part of training data according to various embodiments. [Figure 2B]This figure shows the dependency between the execution time and the number of records for a batch job according to various embodiments. [Figure 2C] This figure shows the dependency between the execution time and the number of records for a batch job according to various embodiments. [Figure 2D] This figure shows batch job execution graphs according to various embodiments. [Figure 2E] This figure shows a portion of the training data according to one embodiment. [Figure 2F] This figure shows an example of a regression model according to various embodiments. [Figure 2G] This figure shows an example of processing performed by a resource estimation system according to various embodiments. [Figure 3] This figure shows the processes performed by the resource estimation system according to various embodiments. [Figure 4A] This figure shows an example of an operation performed by a resource estimation system according to various embodiments. [Figure 4B] This figure shows an example of an operation performed by a resource estimation system according to various embodiments. [Figure 4C] This figure shows an example of an operation performed by a resource estimation system according to various embodiments. [Figure 4D] This figure shows an example of an operation performed by a resource estimation system according to various embodiments. [Figure 4E] This figure shows an example of an operation performed by a resource estimation system according to various embodiments. [Figure 4F] This figure shows an example of an operation performed by a resource estimation system according to various embodiments. [Figure 5] This figure shows the processes performed by the resource estimation system according to various embodiments. [Figure 6] This is a block diagram of a model training and deployment system according to various embodiments. [Figure 7A]This figure shows the processes performed by the resource estimation system according to various embodiments. [Figure 7B] This figure shows the processes performed by the resource estimation system according to various embodiments. [Figure 8] This is a block diagram showing a pattern for implementing a service-type cloud infrastructure system according to at least one embodiment. [Figure 9] This block diagram shows another pattern for implementing a service-type cloud infrastructure system according to at least one embodiment. [Figure 10] This block diagram shows another pattern for implementing a service-type cloud infrastructure system according to at least one embodiment. [Figure 11] This block diagram shows another pattern for implementing a service-type cloud infrastructure system according to at least one embodiment. [Figure 12] A block diagram showing an exemplary computer system according to at least one embodiment. [Modes for carrying out the invention]

[0030] Detailed explanation In the following description, specific details are provided for illustrative purposes to enable a complete understanding of certain embodiments. However, it will become clear that various embodiments can be realized without these specific details. The drawings and this specification are not intended to be limiting in any way. In this specification, the use of the word “exemplary” means “serving as an example, case, or illustration.” No embodiment or design described as “exemplary” in this specification shall necessarily be construed as being preferable or advantageous to other embodiments or designs.

[0031] I. Introduction This disclosure generally relates to the use of artificial intelligence technologies. More specifically and non-limitingly, the technologies disclosed herein relate to novel methods for using machine learning to estimate cloud resources for running batch jobs.

[0032] In computing, batch processing refers to the execution of a group of programs, or jobs, without user interaction. Many businesses use batch processing to perform relatively large volumes of computer processing. For example, a batch job can be a backend process that runs automatically at a given time. Batch jobs can include business processes (such as payment processing and delinquency handling) as well as housekeeping tasks (such as application aging and application parsing). Further examples of batch jobs are given below.

[0033] Batch jobs can be configured for daily, weekly, monthly, and ad-hoc execution. A batch job has multiple records to be processed within the same pipeline. The time required for a batch job varies, generally increasing or decreasing depending on the number of records.

[0034] For some businesses, due to the nature of the tasks, it is not possible to perform and / or complete parts of a task in real time. For example, in the banking industry, there are tasks that can only be completed at the end of a specific day, such as quarterly interest calculations. Such tasks, for example, tasks that must be completed within a certain time frame, are performed by batch processing. Therefore, for some businesses, batch processing is a time-dependent process that requires completion within a specific time frame before a deadline, for example, after business hours but before midnight.

[0035] Due to the time-dependent nature of the process, operators try to ensure sufficient computing resources are available at a given time. For example, interest calculations for a large number of customers (e.g., 1 million to 7 million, etc.) may be performed at the end of a quarter (e.g., the end of March 31), and the interest needs to be displayed for each customer account the following day (e.g., the morning of April 1).

[0036] However, it is difficult to meet batch processing requirements with sufficient available resources without allocating excessive resources. This increases the burden on the computer network, leading to overload and wasted computing resources.

[0037] The technologies described herein overcome the aforementioned problems and other problems. The disclosed technology uses machine learning to ensure that batch jobs are executed within their allocated completion time, reducing the burden on the computer network and conserving computing resources, by using optimal computing resources selected with high precision while avoiding excess resources.

[0038] Because the required computing resources can be temporary and fluctuate based on the number of records being processed, the disclosure method operates in a cloud-based environment. Cloud computing, such as Infrastructure-as-a-Service (IaaS), allows customers to acquire computing resources (in the form of virtual machines) on demand, resulting in computational elasticity.

[0039] The disclosed technology uses machine learning to optimize the allocation of computing resources, finding the most efficient resources for running batch jobs. The most efficient resources are considered to be the fastest and most economical resources capable of running batch jobs within the maximum completion time. Therefore, the disclosed technology improves the functionality of computer networks by conserving computing resources.

[0040] The disclosed technology prevents overloading the computer network by using machine learning to determine the number of virtual machines required to run a job, enabling timely and cost-effective execution of "end-of-day jobs."

[0041] In a cloud environment, the cost of computing machines varies depending on other factors such as time and geographical location. The disclosed technology can determine, based on a job, the number of threads, the number of virtual machines required to run those threads, and the geographical location or data center where the virtual machines should reside so that the job can run in real time within the required timeframe without consuming unnecessary computing resources.

[0042] In various embodiments, the described technology may estimate the number of threads required to run a batch job within a maximum completion time, and use the thread estimate to select the optimal virtual machine or optimal combination of virtual machines from a list of available cloud virtual machines to run the batch job within the maximum completion time.

[0043] In various embodiments, the described technology may use machine learning (ML) models to estimate the number of threads required to run a batch job, and multiple ML models for different types of batch jobs may be constructed using data from historical batch jobs.

[0044] In various embodiments, the described technology allows batch jobs to be executed in real time using an optimal virtual machine whose availability is updated in near real-time.

[0045] As a result of processing using the described technology, a given batch job can be reliably executed within the maximum completion time, while the batch job is completed on time without the need to over-utilize computing resources or to start allocating additional computing resources during the execution of the batch job, thanks to the use of computing resources with maximum efficiency.

[0046] II. Resource Estimation System Figure 1A is a simplified block diagram of a resource estimation system 100 according to one embodiment. The resource estimation system 100 may be implemented using one or more computer systems, each of which has one or more processors. The resource estimation system 100 may comprise a plurality of components and subsystems that are communicated and coupled to one another via one or more communication mechanisms. For example, in the embodiment shown in Figure 1A, the resource estimation system 100 comprises a model generation subsystem 102, a thread prediction subsystem 103, and a virtual machine (VM) selection subsystem 104.

[0047] These subsystems may be implemented as one or more computer systems. The systems, subsystems, and other components shown in Figure 1A may be implemented using software (e.g., code, instructions, programs), hardware, or a combination thereof, executed by one or more processing units (e.g., processors, cores) of each system. The software may be stored in non-temporary storage media (e.g., memory devices).

[0048] The resource estimation system 100 shown in Figure 1A is merely an example and is not intended to unduly limit the scope of embodiments. Many variations, substitutions, and improvements are possible. For example, in some embodiments, the resource estimation system 100 may have more or fewer subsystems or components than shown in Figure 1A, or it may combine two or more subsystems, or the configuration or arrangement of subsystems may differ. The resource estimation system 100 and subsystems shown in Figure 1A may be implemented using one or more computer systems, such as the computer system shown in Figure 12.

[0049] Furthermore, as shown in Figure 1A, the resource estimation system 100 includes a storage subsystem 106 capable of storing various data configurations and programs used by the resource estimation system 100. For example, the storage subsystem 106 may store training data 108 used by the resource estimation system 100, such as historical batch job data. However, this is not intended to be a limitation. For example, the training data 108 may be stored in another storage location, either inside or outside the resource estimation system 100.

[0050] In addition to the training data 108, other data used by or generated by the resource estimation system 100 as part of the functionality of the resource estimation system 100 may be stored in the storage subsystem 106. For example, the cloud virtual machine list 109 may also be stored in the storage subsystem 106.

[0051] As shown in Figure 1B, the resource estimation system 100 can also be provided as part of a distributed computing environment, and the resource estimation system 100 is connected to one or more server computers 126 via a communication network 128. An example of a distributed computing environment is shown in Figure 12 and described in detail below.

[0052] As shown in Figure 1C, the resource estimation system 100 may be part of the CSP infrastructure 129 provided by the CSP to provide one or more cloud services. For example, one or more cloud services may include ABC cloud service 131 to XYZ cloud service 134 connected to the computers of one or more users 135 via a communication network 137. For example, the resource estimation system 100 may be part of ABC cloud service 131.

[0053] An example of the cloud infrastructure architecture provided by CSP is shown in Figures 8-11 and described in detail below.

[0054] In one embodiment, the resource estimation system 100 can determine the number of threads, the number of virtual machines required to execute the threads, and the geographical area or data center where the virtual machines should be located so that the job can be executed in real time, based on the batch job. In one embodiment, the resource estimation system 100 performs multi-stage processing. In the first processing stage, the resource estimation system 100 is configured to estimate the number of threads required to execute the batch job within a given maximum completion time. In the second processing stage, the resource estimation system 100 is configured to predict the optimal virtual machine or optimal combination of virtual machines to process a given batch job.

[0055] A. Thread estimation As detailed below, the thread prediction subsystem 103 estimates the number of threads that can complete a batch job within a time equal to or less than the maximum completion time. However, given a maximum completion time, there may be multiple sets of threads that can satisfy the constraint of completing the batch job within that time.

[0056] For example, if the number of threads is X, a batch job can be completed within Y time, provided that Y is less than or equal to the maximum completion time. If the number of threads is greater than X, a batch job can be completed in less than Y time.

[0057] Each pair of thread counts and corresponding completion times are constraints that serve as inputs to the virtual machine selection subsystem 104. Therefore, the objective of the virtual machine selection subsystem 104 is to find the most efficient virtual machine or combination of virtual machines, given the minimum number of cores as input. In some embodiments, the most efficient virtual machine or combination of virtual machines may represent the most economical solution, e.g., the lowest cost solution. In some embodiments, the most efficient virtual machine or combination of virtual machines may represent the fastest solution. In some embodiments, the most efficient virtual machine or combination of virtual machines may represent the fastest and most economical solution.

[0058] 1. Generating the training dataset In one embodiment, the model generation subsystem 102 may comprise a training dataset generation subsystem 110 and a model training subsystem 114. The training dataset generation subsystem 110 is configured to receive training data 108 as input and generate one or more training datasets by performing certain processing on the training data 108. Each training dataset may contain multiple training data points, each of which is associated with the number of records processed and the completion time for a particular type of historical batch job. The training datasets may be used by the model training subsystem 114 to generate one or more machine learning (ML) models 116.

[0059] In some embodiments, the training dataset generation subsystem 110 may obtain training data 108 from the storage subsystem 106 or another storage location (e.g., an external storage medium accessible to the resource estimation system 100). The training data 108 may be a collection of labeled historical data that provides information about resources used to process multiple historical batch jobs of different types. For example, each historical batch job may be associated with a job code (e.g., a job identifier) ​​that may indicate the type of batch job. Multiple historical batch jobs may include one or more types of batch jobs that represent information collected over a period of time corresponding to the periodicity of batch job execution. For example, this period may range from several weeks to several years.

[0060] Examples of batch job types include calculating interest for financial institution customers, tasks that require the system to update all employees' holiday schedules for the following month by the end of the current month, generating and / or printing periodic reports, and calculating total sales for the day.

[0061] For example, the training data 108 may include historical information for each historical batch job, including batch job attributes. For example, batch job attributes may include one or more of the following: number of threads, completion time, number of virtual machines, number of CPUs, geographical area used, cost, etc. In one embodiment, historical batch jobs may be arranged as groups, and each batch job group includes historical batch jobs of the same batch job type, e.g., interest calculation, and having more diverse information, e.g., at least one different batch job attribute. As an example, a batch job group may include two or more batch jobs that differ in the number of records, number of threads, and / or completion time.

[0062] Figure 2A shows an example of a portion of the training data 108 according to various embodiments.

[0063] Figure 2A shows that training data 108 corresponds to a historical batch job with code TXNTIP_BJ_100_01, for example, a historical batch job among several batch jobs that have been executed in the past. The batch job with job code TXNTIP_BJ_100_01 has 147 records processed by 10 threads over 156 seconds. In some embodiments, the total time associated with processing the historical batch job, for example, 156 seconds in the example of Figure 2A, is considered to be the maximum completion time for that historical batch job.

[0064] Furthermore, the training data 108 provides thread-level information, as shown at the bottom of Figure 2A. For example, for each of threads 1 through 10, the training data 108 provides the number of records and the completion time, e.g., runtime.

[0065] In one embodiment, the training dataset generation subsystem 110 may include a filtering subsystem 120. The filtering subsystem 120 is configured to perform a certain process on the training data 108, for example, to remove historical batch jobs that do not have sufficient correlation between the number of records and the completion time. For example, historical batch jobs are considered to have a correlation if the completion time increases as the number of records increases.

[0066] In some embodiments, the filtering subsystem 120 may include a job pruning filter 122. The job pruning filter 122 is configured to receive training data 108 as input, corresponding to each group of historical batch jobs containing historical batch jobs of the same type. The job pruning filter 122 is further configured to determine a correlation coefficient between execution time and the number of records for each batch job type. The job pruning filter 122 is further configured to remove historical batch jobs whose correlation coefficient is less than a correlation threshold. In some embodiments, the correlation threshold is adjustable and may be set to, for example, 0.6.

[0067] Figures 2B and 2C show the distribution of the number of records versus execution time for historical batch jobs according to various embodiments.

[0068] Graph 220 in Figure 2B relates to the first type of historical batch job group, but does not show sufficient correlation. As shown in the figure, the completion times for different numbers of records are approximately the same, meaning that the first type of historical batch job is not processed consistently. Also, some data is missing.

[0069] Graph 222 in Figure 2B relates to the second type of historical batch job group. Similar to Graph 220, this batch job type also does not show sufficient correlation. As shown in the figure, the completion time varies even when the number of records is the same, meaning that the second type of historical batch job is not processed consistently, and therefore Graph 222 does not show sufficient correlation. In addition, only a portion of the data for a small number of records is present.

[0070] Graph 224 in Figure 2C relates to the third type of historical batch job. As shown in the figure, the completion time increases as the number of records increases. In other words, because the historical batch jobs are processed consistently, Graph 224 shows a sufficient correlation.

[0071] Therefore, the job pruning filter 122 of the filtering subsystem 120 can distinguish between groups of historical batch jobs that do not show sufficient correlation and groups of historical batch jobs that do show sufficient correlation in the training data 108 by calculating the correlation coefficient of multiple historical batch jobs of each batch job type. The job pruning filter 122 can exclude one or more historical batch jobs with a correlation coefficient below the correlation threshold from separate processing, and allow historical batch jobs with a correlation coefficient above the correlation threshold to pass to separate processing.

[0072] a. Calculation of thread initialization time and process execution time In one embodiment, the resource estimation system 100 may further comprise a data processor 124. The data processor 124 is configured to receive historical batch jobs that have passed through the job pruning filter 122 as input and to perform certain processing associated with these historical batch jobs. For example, the data processor 124 may calculate thread initialization time and / or process execution time.

[0073] Thread initialization time and process execution time will be discussed later with reference to Figures 2D and 2E.

[0074] Figure 2D shows a batch job execution graph according to one embodiment. Figure 2E shows a portion of the training data 108 indicating the thread start and end times according to one embodiment.

[0075] In the example shown in Figure 2D, a batch job (e.g., a history batch job) has 5,000 records. As shown in Graph 200, the batch job has a first thread 202, a second thread 204, a third thread 206, a fourth thread 208, and a fifth thread 210. The bars in Graph 200 are of equal length. This indicates that the records are equally divided among the first to fifth threads 202 to 210. Since the records in Graph 200 are equally divided among the first to fifth threads 202 to 210, the execution time of each of the first to fifth threads 202 to 210 is approximately the same.

[0076] A batch job starts at time A, corresponding to the job start time, and ends at time B, corresponding to the job completion time.

[0077] The time interval between A and B must be less than or equal to the maximum completion time in order to satisfy the constraint of the maximum completion time. The time interval from A to B consists of three parts.

[0078] For example, the period T1 from point A, when the batch job starts, until the first thread 202 is generated, corresponds to the first part of the time interval from A to B. As shown in Graph 200, each of the periods T2 until the second thread 204 starts, T3 until the third thread 206 starts, T4 until the fourth thread 206 starts, and T5 until the fifth thread 210 starts is shorter than period T1. The sum of periods T2 to T5 corresponds to the second part of the time interval from A to B. These periods T2 to T5 may be equal.

[0079] In the use herein, thread initialization time represents the sum of the first and second parts of time interval A to B.

[0080] The third part of the time interval from A to B corresponds to the process execution time during which the thread is processed. For example, period TE5 corresponds to the process execution time.

[0081] Referring to Figure 2E, the training data 108 may include thread-level data for a historical batch job incorporating the timings described above, referring to Figure 2D. For example, for a historical batch job with code TXNTIP_BJ_100_01, Figure 2E shows thread-level data for five threads, including thread start time, thread end time, and creation time.

[0082] In one embodiment, the data processor 124 can calculate the time until the start of the first thread and the time until the start of the subsequent threads for each history batch job, based on thread-level data. In an example in Figure 2E, the time until the start of the first thread is 98 seconds (4:21:56 to 4:23:34), and the time until the start of the subsequent threads is 2 seconds, for example, 4:23:36 to 4:23:34.

[0083] The data processor 124 can calculate the thread initialization time according to Equation 1. Formula 1: Thread initialization time = [Time until the first thread starts + Time until consecutive threads start × (Total number of threads - 1)] In the example shown in Figure 2E, the thread initialization time is 98 + 2(5-1) = 106 seconds.

[0084] Given the maximum completion time, the data processor 124 can calculate the process execution time according to Equation 2.

[0085] Formula 2: Process execution time = Maximum completion time - Thread initialization time The training dataset generation subsystem 110 may further comprise a training dataset generator 133. The training dataset generator 133 combines the outputs provided by the filtering subsystem 120 and the data processor 124 to generate one or more training datasets corresponding to a certain type of historical batch job. The training datasets may include data corresponding to threads, the number of records in the corresponding historical batch job, thread initialization time, and thread process execution time.

[0086] 2. Model Training The model training subsystem 114 receives an input model 130 of a desired architecture and one or more training datasets as input. In some embodiments, the input model 130 may be a machine learning model.

[0087] In the use herein, “machine learning model” or “model” may represent a software module that runs on one or more processors and is configured to provide classifications or numerical representations of the characteristics of one or more samples. An exemplary model type is supervised learning, which is available in embodiments of this disclosure. Exemplary supervised learning models may include a variety of techniques and algorithms, such as analytical learning, artificial neural networks, backpropagation, boosting (meta-algorithm), Bayesian statistics, case-based reasoning, decision tree learning, inductive logic programming, Gaussian process regression, genetic programming, data processing using group methods, kernel estimators, learning automata, learning classifier systems, minimum message length (decision trees, decision graphs, etc.), multilinear subspace learning, naive Bayes classifiers, maximum entropy classifiers, conditional random fields, nearest neighbor algorithms, probabilistic and approximately correct (PAC) learning, ripple-down rules, knowledge acquisition methods, symbolic machine learning algorithms, subsymbolic machine learning algorithms, minimal complexity machines (MCMs), random forests, classifier ensembles, ordinal classification, statistical relational learning, or the multi-criterion classification algorithm Proaftn.

[0088] The models may include linear regression, logistic regression, deep recurrent neural networks (e.g., long-short-term memory, LSTM), hidden Markov models (HMM), linear discriminant analysis (LDA), k-means clustering, density-based clustering applications with associated noise (DBSCAN), random forest algorithms, support vector machines (SVM), or any model described herein. Supervised learning models can be trained in various ways by using various cost or loss functions that define the error from known labels (e.g., least squares and absolute difference from known classifications) and various optimization techniques, such as backpropagation, steepest descent, conjugate gradient, Newton and quasi-Newton methods.

[0089] In some embodiments, machine learning models may include, but are not limited to, convolutional neural networks (CNNs), linear regression, logistic regression, deep recurrent neural networks (e.g., fully connected recurrent neural networks (RNNs), gated recurrent units (GRUs), long-shortened memory (LSTMs)), transformation-based methods (e.g., XLNet, BERT, XLM, RoBERTa), Bayesian classifiers, hidden Markov models (HMMs), linear discriminant analysis (LDA), k-means clustering, noise associated with density-based clustering applications (DBSCAN), random forest algorithms, adaptive boosting (AdaBoost), extreme gradient boosting (XGBoost), support vector machines (SVMs), or composite models including one or more of the above-proposed models.

[0090] In one embodiment, the architecture of the input model 130 may be one of the architectures that a person skilled in the art would know to be suitable for the task. For example, the input model 130 may be a regression model that can provide a function describing the relationship between one or more independent variables and a response-dependent variable or target variable.

[0091] The model training subsystem 114 generates one or more ML models 116 corresponding to specific types of batch jobs by training the input model 130 using one or more training datasets.

[0092] As a result of training performed by the model training subsystem 114, the generated ML model 116 can be used to determine whether a batch job can be completed within the maximum completion time, given the number of records in the batch job and the maximum completion time as input. If the ML model 116 determines that the batch job can be completed within the maximum completion time, it can determine, for example, predict or estimate the number of threads required to process the records of the batch job within the maximum completion time. Subsequently, the ML model 116 provides the number of threads and the corresponding completion time for the given batch job as output. The processing performed by the ML model 116 is described in detail below.

[0093] Figure 2F shows an example of the regression model 300 (e.g., a regression graph) for the history batch job type in Figure 2C. The regression model 300 may correspond to the ML model 116. Although Figure 2F shows a linear regression, no limitation is intended. In some embodiments, the regression may be nonlinear.

[0094] For the sake of simplicity, the embodiment is described for the generation of ML model 116 for a single type of batch job. However, additional ML models 116 may be generated for different batch job types as well.

[0095] In one embodiment, each ML model 116 generated by the model training subsystem 114 may be stored in the storage subsystem 106, for example, associated with a job code corresponding to a batch job type. The time until the start of the first thread and the time until the start of the successive threads may also be stored, respectively, in association with the ML model 116.

[0096] Model training and deployment are described in more detail in Figure 6.

[0097] 3. Thread prediction As described above, the resource estimation system 100 may further comprise a thread prediction subsystem 103. The thread prediction subsystem 103 may receive batch job data as input, including the job code, number of records, and the associated maximum completion time for the batch job, and use the corresponding ML model 116 along with the relevant information to predict the number of threads needed to process the batch job within the maximum completion time.

[0098] In an embodiment, the thread prediction subsystem 103 may determine (e.g., predict or estimate) the number of threads required to process a given number of records in a batch job and the actual completion time within the maximum completion time by using a thread estimation algorithm based on the maximum completion time and the number of records. The thread estimation algorithm can predict the number of threads and the actual completion time using an ML model 116 corresponding to the job code (e.g., the batch job type of a batch job). An example and description of the thread estimation algorithm are given below.

[0099] a. Thread estimation algorithm using regression models [Initializing variables] List_threads_completiontime = [] max_thread = maximum allowed threads passed as configuration # For example, 1000 total_records = number of records # For example, 25335 max_completion_time = maximum time to complete the job # For example, 1600 seconds [Algorithm pseudocode] Iterate no_thread from 2 ... max_thread Step-1 : Calculate thread_initialization_time Step-2 : process_execution_time = max_completion_time - thread_initialization_time Step-3 : record_processed_predict = regression_1 (process_execution_time) Step-4 : record_processed_expected = total_records / no_thread Step-5 : If (record_processed_predict >= record_processed_expected) then a)actual_completion_time = regression_2 (record_processed_expected) + thread_initialization_time b)add (no_thread, actual_completion_time) to List_threads_completiontime Return List_threads_completiontime After the variables are initialized, the thread estimation algorithm is executed iteratively for different numbers of threads, such as 2, 3, 4, ..., 10, ..., 100, ..., 1000, etc. In an unrestricted example, the number of threads may be a natural number from 2 to a predetermined maximum number, such as 1000.

[0100] In step 1, for each number of threads, the data processor 124 may determine the thread initialization time for the same job code based on the time to start of the first thread and the time to start of subsequent threads previously stored in association with the ML model 116 trained for that job code, for example. However, this is not intended to be any limitation. In some embodiments, the batch job data may include thread-based data as described above with reference to Figure 2E. In such embodiments, the data processor 124 may calculate the time to start of the first thread and the time to start of subsequent threads, and then calculate the thread initialization time as described above.

[0101] In step 2, the data processor 124 may calculate the process execution time by subtracting the thread initialization time from the maximum completion time according to equation 2.

[0102] In step 3, the ML model 116 receives the process execution time as input and, for example, determines the predicted number of records that are expected to be processed by the current number of threads within the process execution time for the current iteration, which it provides as output.

[0103] In step 4, the thread prediction subsystem 103 can determine the expected number of records that are expected to be processed by the current number of threads by dividing the total number of records in the batch job by the current number of threads in the current iteration.

[0104] In step 5, if the predicted number of records (for example, obtained as the output of ML model 116) is greater than or equal to the expected number of records, the thread prediction subsystem 103 provides the expected number of records as input to ML model 116. The ML model 116 can then output, for example, the actual completion time for processing the expected number of records in the current iteration. The thread prediction subsystem 103 then stores the data of the current iteration, such as the actual completion time output by ML model 116, the expected number of records to be processed, the thread initialization time, and the number of threads in the current iteration, in the database 138. However, the above is not intended to be limiting in any way. For example, the thread prediction subsystem 103 can store the data of the current iteration in the storage subsystem 106.

[0105] As a result of multiple iterations, the thread prediction subsystem 103 may obtain one or more sets of data items, including the number of threads and the corresponding predicted completion time.

[0106] For example, the thread prediction subsystem 103 stores the iteration data in the database 138 as a thread and completion time list 140, for example, a table, which provides a correspondence between actual completion time and the number of threads.

[0107] In one embodiment, the thread and completion time list 140 may be displayed to the user or output to an external device.

[0108] b. Example of batch job data processing using a thread prediction algorithm Figure 2G shows examples of possible batch job data for a batch job with a given job code. For simplicity of explanation, possible batch job data may be referred to as test batch job data. As shown by reference number 310, batch jobs of the same type (for example, having the same job code) may have different batch job data. In one example, the test batch job data has three examples: (A) number of records = 84072, maximum completion time = 3000 seconds, (B) number of records = 10013, maximum completion time = 500 seconds, and (C) number of records = 164264, maximum completion time = 800 seconds.

[0109] In Figure 2G, reference number 312 indicates the output provided by the thread prediction subsystem 103. For example, for test batch job data (A) having 84072 records, the thread prediction subsystem 103, for example, the thread estimation algorithm, may output three possible solutions, for example, combinations of thread count and completion time: combination (1) having a first data item corresponding to the thread count (128) and a second data item corresponding to the completion time (2990 seconds); combination (2) having a first data item corresponding to the thread count (150) and a second data item corresponding to the completion time (2800 seconds); and combination (3) having a first data item corresponding to the thread count (256) and a second data item corresponding to the completion time (1800 seconds). The thread and completion time list 140 described above may include, for example, one combination for test batch job data (B), or it may include multiple combinations similar to the combinations for test batch job data (A).

[0110] 4. Thread estimation method Figure 3 shows processing according to various embodiments. For example, the processing 320 shown in Figure 3 may be executed by some or all of the training dataset generation subsystem 110, the model training subsystem 114, the data processor 124, and the thread prediction subsystem 103.

[0111] The process 320 shown in Figure 3 may be implemented by the use of software (e.g., code, instructions, programs), hardware, or a combination thereof, executed by one or more processing units (e.g., processors, cores) of each subsystem. The software may be stored in a non-temporary storage medium (e.g., a memory device). The methods presented in Figure 3 and described below are illustrative and not intended to be limiting. Figure 3 shows various processing operations occurring in a particular order or sequence, but not intended to be limiting. In an alternative embodiment, the process 320 may be executed in some different order, or at least some of the operations may be executed in parallel.

[0112] In operation 322, the training dataset generation subsystem 110, for example, the filtering subsystem 120 or the job pruning filter 122, receives training data 108 corresponding to historical batch jobs as input and is configured to find a correlation coefficient between the execution time of the batch jobs and the number of records. The training dataset generation subsystem 110 may be configured to remove historical batch jobs having a correlation coefficient below a correlation threshold, as described above.

[0113] In operation 324, the data processor 124 receives filtered historical batch jobs as input and is configured to calculate the thread initialization time and process execution time of the historical batch jobs as described above.

[0114] In operation 326, the model training subsystem 114 receives the training dataset output from the training dataset generation subsystem 110. As detailed above, the model training subsystem 114 uses the training dataset to perform model training and generate the ML model 116.

[0115] In operation 328, the thread prediction subsystem 103 may receive batch job data 132 as input, which includes the number of records in the batch job and the associated maximum completion time. As described above, the thread prediction subsystem 103 may determine (e.g., predict or estimate) the number of threads required to process the records of the batch job and the actual completion time within the maximum completion time by using the thread estimation algorithm and the ML model 116 based on the maximum completion time and the number of records.

[0116] In operation 330, the thread prediction subsystem 103 may store records corresponding to batch jobs in list 140, as described above. Each record may include an entry that corresponds to the number of threads in the batch job and includes the actual completion time.

[0117] B. Selecting a virtual machine The output of the thread prediction subsystem 103 is passed to the virtual machine selection subsystem 104, which performs the processing corresponding to the second processing stage. For example, the output of the thread prediction subsystem 103 may include the actual completion time and number of threads for a given batch job stored in the thread and completion time list 140. The received information may include a data column containing a first data item corresponding to the number of threads and a second data item corresponding to the completion time, or a data column containing two or more combinations of the first and second data items, for example, {th1, T1, th2, T2, ...thn, Tn}. The first data items th1 to thn correspond to different numbers of threads, for example, 2, 10, ..., 1000. The second data items T1 to Tn correspond to the completion time corresponding to the number of threads. The virtual machine selection subsystem 104 is configured to determine the optimal single virtual machine or optimal combination of virtual machines to process the batch job by performing an iterative operation for each pair of the first and second data items.

[0118] In one embodiment, the virtual machine selection subsystem 104 comprises a virtual machine selector 150. The virtual machine selector 150 is configured to receive data from a thread and completion time list 140 as input from the thread prediction subsystem 103, which includes the number of threads (th) and actual completion time (T) derived for batch jobs that should be completed within a maximum completion time. The virtual machine selector 150 can calculate the number of cores required for a batch job using Equation 3.

[0119] Formula 3: Number of cores = Number of threads / Number of threads per core Here, the number of threads per core may be predetermined, or in some embodiments, it may be equal to 2.

[0120] Therefore, for example, if the virtual machine selector 150 receives a value of 80 as the number of threads (th), the number of cores can be calculated as equal to 40 (80 / 2). For example, a minimum of 40 cores may be required to run a given batch job within the maximum completion time.

[0121] Furthermore, the virtual machine selector 150 can receive information about available cloud virtual machines. This information may include the geographical region where a particular virtual machine is located, the hourly usage cost, and the core configuration (e.g., number of cores).

[0122] In one example, the number of cores may be 18, 32, 64, or 80, and each core configuration may have different associated costs for the same or different geographical regions. However, there is no limit to the number of cores, and any appropriate number may be used.

[0123] For example, real-time or near-real-time information about available cloud virtual machines may be stored in the cloud virtual machine list 109. In various embodiments, the cloud virtual machine list 109 may be updated periodically. In one embodiment, the cloud virtual machine list 109 may be self-updated using a Cron job by calling an API, for example, every hour or at any other preferred time interval.

[0124] The virtual machine selector 150 may be configured to determine one or more virtual machines (for example, a single virtual machine or a combination of virtual machines) having 40 or more cores.

[0125] In one embodiment, the virtual machine selector 150 selects the most efficient single virtual machine from among virtual machines having the same number of cores, for example, 18, 32, 64, or 80, in each category. The optimal single virtual machine may have the lowest associated cost.

[0126] Figures 4A to 4F show examples of operations performed by the resource estimation system 100 according to various embodiments.

[0127] In some embodiments, the virtual machine selector 150 first sorts single virtual machines from the cloud virtual machine list 109 by the number of OCPUs. Then, single virtual machines having the same number of OCPUs may be grouped together as the same group.

[0128] Figure 4A shows an example of a list 400, for example, a group containing single virtual machines with the same number of OCPUs, for example, 64. Similarly, the virtual machine selector 150 may configure groups having single virtual machines with 18, 32, and 80 OCPUs. The virtual machine selector 150 can then select the most efficient single virtual machine from each group.

[0129] As shown by reference no. 402 in Figure 4A, the virtual machine selector 150 can select a virtual machine located in Sydney, Australia, as the most efficient single virtual machine among the single virtual machines on list 400. In a non-limiting example, the most efficient single virtual machine is the most economical, for example, the one with the lowest cost, among a group of single virtual machines with the same number of OCPUs.

[0130] Figure 4B shows a list of 404 maximum-efficiency single virtual machines selected by the virtual machine selector 150 in each category from all geographical regions of the cloud virtual machine list 109, for example, each having 18, 32, 64, or 80 OCPUs.

[0131] Optionally, the virtual machine selector 150 can store a list of 404 single virtual machines with maximum efficiency.

[0132] The virtual machine prediction subsystem 160 is configured to predict the optimal single virtual machine or optimal combination of virtual machines for executing a batch job, for example, by running a virtual machine prediction algorithm on the most efficient single virtual machine from List 404.

[0133] Virtual machine prediction algorithm [Initializing variables] List_economical_vm = [] CPU = List of available CPU cores # For example, [18, 32, 64, 80] Cost = List of costs with respect to available CPU cores #For example, [13, 15, 20, 44] List_total_cores = List of possible cores which allows the job to execute in max_completion_time [Algorithm pseudocode] Iterate total_cores from List_total_cores Recursion (current_cores= 0, index = 0, current_cost = 0, current_vm = []) if (current_cores >= total_cores) then add [current_cores, current_cost, current_vm] to List_economical_vm if (current_cores >= total_cores) or (index = length(CPU)) then Return recursion (current_cores += CPU[index], index, current_cost += Cost[index], current_vm + [CPU[index]]) # The current virtual machine is selected recursion(current_cores, index + 1, current_cost, current_vm) # The current virtual machine is not selected. Return List_economical_vm For example, by executing a virtual machine prediction algorithm, the virtual machine prediction subsystem 160 can determine, for example, predict, which high-efficiency virtual resources should be included in a set of virtual machines. Each high-efficiency virtual resource may include a single virtual machine or a combination of virtual machines. The high-efficiency virtual resources may be arranged in a list of high-efficiency virtual resources in descending order of efficiency, for example, ascending order of cost, for a given combination of a first data item and a second data item.

[0134] The virtual machine prediction algorithm may be executed iteratively for each pair of thread count and completion time, and the results are predictions of a single virtual machine, multiple single virtual machines, a combination of virtual machines, and / or multiple combinations of virtual machines. After each iteration, the virtual machine prediction subsystem 160 may place the results of the iteration in a high-efficiency virtual resource list in ascending order of cost. For example, the virtual machine prediction subsystem 160 may generate multiple high-efficiency virtual resource lists for multiple iterations for each pair of thread count and completion time, as shown in Figures 4E and 4F and described below. However, if there is only one execution, the virtual machine prediction subsystem 160 generates one high-efficiency virtual resource list, as shown in Figure 4D and described below.

[0135] In one embodiment, one or more highly efficient virtual resource lists may be displayed on a display or output to an external device.

[0136] The virtual machine prediction subsystem 160 may also select the best virtual machine 170 with the lowest cost (for example, the best single virtual machine or the best combination of virtual machines) from among the highly efficient virtual resources in one or more highly efficient virtual resource lists.

[0137] This optimized virtual machine 170 may be used to run batch jobs, such as real-time batch jobs.

[0138] Figure 4C shows the process 420 performed by the virtual machine prediction subsystem 160, which executes the virtual machine prediction algorithm shown above.

[0139] Reference number 422 indicates the start of execution of the virtual machine prediction algorithm. As shown in the diagram, total cores = 0 and cost = 0.

[0140] Next, the virtual machines in List 404, shown in Figure 4B, are selected recursively, with cores and costs added in each iteration. The shaded boxes in Figure 4C indicate that the combination of virtual machines resulted in a core count of 40 or more. For example, in box 424, the core count is 50 and the cost is 28. Box 424 shows a derived combination that includes an 18-core virtual machine in Bangalore with metric 0 and a 32-core virtual machine in Sydney with metric 1 (see Figure 4B).

[0141] As shown in Figure 4D, the virtual machine prediction subsystem 160 may use the information in the shaded box in Figure 4C to generate a highly efficient virtual resource table 430, for example, using information on the total number of cores greater than 40 and the cost associated with each solution. The virtual machine prediction subsystem 160 may sort the virtual resources based on cost, for example, by arranging single virtual machines or combinations of virtual machines in ascending order of cost. The virtual machine prediction subsystem 160 may also include a predetermined number of the most economical virtual machines in an ordered sequence. The virtual machine prediction subsystem 160 may then select the optimal virtual machine or optimal combination of virtual machines from the ordered sequence that satisfies the number of cores required to run a batch job within the maximum completion time and has the lowest cost.

[0142] Referring again to Figure 4D and then to Figure 4B, the virtual machine prediction subsystem 160 can select a virtual machine of shape "VM.Standard.E3.flex" located in Sydney and having 64 cores as the optimal virtual machine from Table 430.

[0143] The example above represents, for instance, a single iteration concerning the number of threads and completion time for a batch job. However, the virtual machine selection subsystem 104 performs similar processing for multiple combinations of thread count and completion time.

[0144] In an example described later with reference to Figure 4E, the thread prediction subsystem 103 provides a list of threads and completion times 140 {th1=600, t1=200 seconds, th2=900, t2=140 seconds} as input to the virtual machine selection subsystem 104. The maximum completion time is 210 seconds.

[0145] The virtual machine selector 150 can calculate the required number of cores as 300 (300 / 2) if the completion time is 200 seconds, and 450 (900 / 2) if the completion time is 140 seconds.

[0146] Figure 4E shows Table 440 obtained as a result of processing performed by the virtual machine selector 150 and the virtual machine prediction subsystem 160 for 300 cores and a completion time of 200 seconds.

[0147] Figure 4F shows Table 442, obtained as a result of processing performed by the virtual machine selector 150 and the virtual machine prediction subsystem 160 for 450 cores and a completion time of 140 seconds.

[0148] Tables 440 and 442 are generated as a result of iteratively running the virtual machine prediction algorithm on the combinations {th1=600, t1=200 and th1=900, t1=140}, respectively.

[0149] Based on the results obtained in Tables 440 and 442, the virtual machine prediction subsystem 160 may select the optimal virtual machine combination from Table 442, for example, a single 18-core virtual machine of shape "VM.Optimized.E3.flex" located in Bangalore and seven 64-core virtual machines of shape "VM.Standard.E3.flex" located in Sydney (see Figure 4B).

[0150] Therefore, in the above example, the fastest virtual machine combination is selected, which is also the most economical.

[0151] 1. Virtual Machine Prediction and Selection Method Figure 5 shows processing according to various embodiments. For example, at least a portion of the processing 500 shown in Figure 5 may be executed by the virtual machine selection subsystem 104.

[0152] The process 500 shown in Figure 5 may be implemented by one or more processing units of each subsystem, such as a processor, software executed by a core, such as code, instructions, programs, hardware, or a combination thereof. The software may be stored in a non-temporary storage medium (e.g., a memory device). The methods presented in Figure 5 and described below are illustrative and not intended to be limiting. Figure 5 shows various processing operations occurring in a particular order or sequence, but not intended to be limiting. In an alternative embodiment, the process 500 may be executed in some different order, or at least some of the operations may be executed in parallel.

[0153] In operation 501, the virtual machine selector 150 may receive real-time or near-real-time information about available cloud virtual machines that may be stored in the cloud virtual machine list 109. The virtual machine selector 150 can then perform the process of selecting the most efficient single virtual machine in each category, for example, the most efficient single virtual machines having 18, 32, 64, and 80 OCPUs, respectively, as described above with reference to Figures 4A and 4B.

[0154] In operation 502, the virtual machine selection subsystem 104 may receive information from the model generation subsystem 102, such as the number of threads and the completion time, as described above. The received information may include a data sequence containing a first data item corresponding to the number of threads and a second data item corresponding to the completion time, or a data sequence containing two or more combinations of the first data item and the second data item, for example, {th1, T1, th2, T2, ...thn, Tn}. The first data items th1 to thn correspond to different numbers of threads, for example, 2, 10, ..., 1000. The second data items T1 to Tn correspond to the completion time corresponding to the number of threads. The virtual machine selection subsystem 104 is configured to perform an iterative operation for each pair of the first and second data items.

[0155] In operation 504, the virtual machine prediction subsystem 160 can run the virtual machine prediction algorithm for the single most efficient virtual machine.

[0156] In operation 506, the virtual machine prediction subsystem 160 can determine, for example, predict, which high-efficiency virtual resources should be included in the virtual machine set. High-efficiency virtual resources may include at least one of a single virtual machine, multiple single virtual machines, a combination of virtual machines, and / or multiple combinations of virtual machines. High-efficiency virtual resources may be placed in the high-efficiency virtual resource list in ascending order of cost for each pair of first and second data items, for example, thread count and completion time.

[0157] Operations 504 and 506 are described in more detail above with reference to Figures 4C to 4F. In an embodiment, operations 504 and 506 may be executed iteratively for each pair of thread count and completion time, and the results are predictions of a single virtual machine, multiple single virtual machines, a combination of virtual machines, and / or multiple combinations of virtual machines. After each iteration, the virtual machine prediction subsystem 160 may place the results of the iteration in a high-efficiency virtual resource list in ascending order of cost. For example, the virtual machine prediction subsystem 160 may generate multiple high-efficiency virtual resource lists for multiple iterations, as shown in Figures 4E and 4F. However, if operations 504 and 506 are executed only once, the virtual machine prediction subsystem 160 generates one high-efficiency virtual resource list, as shown in Figure 4D.

[0158] In operation 508, the virtual machine prediction subsystem 160 may select the optimal virtual machine (e.g., a single virtual machine) or optimal virtual machine combination with the lowest cost from one list of highly efficient virtual resources or multiple lists of highly efficient virtual resources.

[0159] In operation 518, a batch job, such as a real-time batch job, may be executed using the optimal virtual machine or optimal combination of virtual machines selected in operation 508.

[0160] III. Model Training and Deployment Figure 6 is a block diagram showing a machine learning system 600 according to various embodiments. For example, the machine learning system 600 may be part of the resource estimation system 100, or it may be in communication with the resource estimation system 100, thereby facilitating model training.

[0161] As shown in Figure 6, the machine learning system 600 includes various stages: a predictive model training stage 610 for building and training models; an evaluation stage 215 for evaluating the performance of the trained models; and an implementation stage 620 for implementing one or more models. The predictive model training stage 610 builds and trains one or more predictive models 225a and 225b to 225n (where "n" is any natural number) (which may be referred to individually or collectively as predictive models 225 in this specification) used in the other stages. For example, predictive models 225 may include any machine learning models described above with respect to the resource estimation system 100.

[0162] The prediction model 225 may be a machine learning model of the type described above. The resource estimation system 100 may employ the same or different types of prediction models to provide prediction results to the user. In other examples of the embodiment, yet other types of prediction models may be implemented.

[0163] To train various predictive models 225, the predictive model training phase 610 may include a dataset preparation module 230, a model training framework 240, and an evaluation phase 215. The dataset preparation module 230 performs operations such as loading data assets 245 (e.g., training datasets), splitting data assets 245 into training and validation sets 245a to n so that the system can train and test predictive models 225, and preprocessing data assets 245. The splitting of data assets 245 into training and validation sets 245a to n may be performed randomly (e.g., 60 / 40%, 70 / 30%, etc.).

[0164] The model training framework 240 performs the processes of determining the hyperparameters of the predictive model 225 and performing iterative operations to input examples from the training data 245a into the predictive model 225 to find a set of model parameters (e.g., weights and / or biases) that minimize the cost function, such as the loss function or error function, of the predictive model 225. Hyperparameters are settings that can control the behavior of the predictive model 225 through adjustment or optimization. Most models explicitly define hyperparameters that control various aspects of the model, such as memory or execution cost. However, additional hyperparameters may be defined to adapt the predictive model 225 to specific scenarios, such as learning rate, number of iterations, regularization weights, or intensity.

[0165] The cost function may be configured to measure the difference between the output inferred using the predictive model 225 and the ground truth annotated with labels for the samples. For example, in a supervised learning-based model, the goal of training is to learn a function "h()" (sometimes called the hypothesis function) h:X→Y that maps the training input space X to the target value space Y, where h(x) is a good predictor for the corresponding value of y. Various different techniques may be used to learn this hypothesis function. In some techniques, as part of the derivation of the hypothesis function, a cost function or loss function may be defined that measures the difference between the ground truth value for an input and the predicted value for that input. As part of the training, this cost function or loss function is minimized by using techniques such as backpropagation, random feedback, direct feedback alignment (DFA), indirect feedback alignment (IFA), and Hebb learning.

[0166] Once a set of model parameters is identified, the model 225 is trained, and the model training framework 240 performs an additional testing or validation process using a subset of the test data 245b (test or validation dataset). The testing or validation process involves iterative operations to find the optimal set of hyperparameters by tuning the hyperparameters by inputting samples from the subset of test data 245b into the predictive model 225 using validation techniques such as K-Fold Cross-Validation, Leave-one-out Cross-Validation, Leave-one-group-out Cross-Validation, and Nested Cross-Validation. Once the optimal set of hyperparameters is obtained, the test dataset secured from the subset of training data 245a is input into the predictive model 225 to obtain an output, and the output is evaluated against ground truth using correlation techniques such as Brand-Altman and Spearman's rank correlation coefficient. Furthermore, in the evaluation stage 215, performance criteria 255 such as error, precision, accuracy, recall, and receiver operating characteristic curve (ROC) may be calculated. In evaluation stage 215, performance criteria 255 may be used to analyze the performance of the predictive model 225.

[0167] The prediction model training stage 610 outputs a trained model containing one or more trained models 260. One or more trained models 260 may correspond to one or more ML models 116. One or more trained models 260 can be deployed and used in the implementation stage 620 to provide prediction results 265 as described above. For example, a trained model 260 can take input data 270 (e.g., number of records, completion time, etc.) and provide prediction results (or output) 265 to the user.

[0168] IV. Method Figure 7A shows a process according to various embodiments. For example, the process 700 shown in Figure 7A may be executed by part or all of the resource estimation system 100, such as the model generation subsystem 102, the thread prediction subsystem 103, and the virtual machine selection subsystem 104.

[0169] The process 700 shown in Figure 7A may be implemented by the use of software (e.g., code, instructions, programs), hardware, or a combination thereof, executed by one or more processing units (e.g., processors, cores) of each subsystem. The software may be stored in a non-temporary storage medium (e.g., a memory device). The methods presented in Figure 7A and described below are illustrative and not intended to be limiting. Figure 7A shows various processing operations occurring in a particular order or sequence, but not intended to be limiting. In an alternative embodiment, the process 700 may be executed in some different order, or at least some of the operations may be executed in parallel.

[0170] Continuing to refer to Figure 7A, in operation 702, the resource estimation system 100 may estimate the number of threads that will execute a batch job with a total number of records within the maximum completion time by executing a thread estimation algorithm. For example, the resource estimation system 100 may use a machine learning (ML) model to determine one or more combinations of thread estimates, each including the estimated number of threads and the actual completion time for processing the total number of records using the estimated number of threads, wherein the actual completion time is less than or equal to the maximum completion time.

[0171] The resource estimation system 100 may iteratively execute a thread estimation algorithm for each of several thread counts between 2 and a predetermined maximum number of threads. For example, the resource estimation system 100 may determine the thread initialization time based on the current number of threads for processing a batch job in the current iteration of several iterations, determine the process execution time using the maximum completion time and the thread initialization time, provide the process execution time as input to an ML model, obtain the predicted number of records to be processed by the current number of threads within the process execution time as output to the ML model, determine whether the predicted number of records is greater than or equal to the expected number of records that can be processed using the current number of threads, and if the predicted number of records is greater than or equal to the expected number of records, use the ML model to determine the actual completion time for processing the expected number of records, associate the current number of threads with the actual completion time of the batch job, and save the current number of threads corresponding to the estimated number of threads in the current iteration in a thread and completion time list.

[0172] The resource estimation system 100 may use an ML model to determine the actual completion time for processing the expected number of records by providing the expected number of records as input to the ML model. The expected number of records to be processed in the current iteration is calculated by dividing the total number of records by the current number of threads. The execution time corresponding to the expected number of records is obtained as the output of the ML model, and the actual completion time is calculated by adding the thread initialization time to the execution time output by the ML model.

[0173] Subsequently, the resource estimation system 100 may output a list of threads and their completion times based on multiple iterations performed for multiple threads.

[0174] The resource estimation system 100 may generate a training dataset using training data associated with a certain type of historical batch job, and then generate an ML model for processing a certain type of batch job by training an input model using the training dataset.

[0175] Similarly, the resource estimation system 100 may generate multiple training datasets using multiple sets of training data associated with a certain type of historical batch job among multiple batch job types. The resource estimation system 100 may also generate multiple ML models for processing batch jobs corresponding to multiple batch job types by training input models using the multiple training datasets.

[0176] ML models include regression models. In operation 704, the resource estimation system 100 may store a set containing one or more combinations of threads and completion time lists.

[0177] In operation 708, the resource estimation system 100 may select an optimal virtual machine or optimal virtual machine combination from a list of available cloud virtual machines in multiple geographical areas, using a list of threads and completion times, the optimal virtual machine or optimal virtual machine combination being able to execute batch jobs within the maximum completion time.

[0178] The resource estimation system 100 may select multiple maximum efficiency single virtual machines from the cloud virtual machine list, each of which corresponds to a number of cores per single virtual machine, and the resource estimation system 100 may place the multiple maximum efficiency single virtual machines in the list of the multiple maximum efficiency single virtual machines.

[0179] The thread and completion time lists may include multiple combinations, each containing an estimated number of threads and an actual completion time. The resource estimation system 100 may select the optimal virtual machine or optimal virtual machine combination by iterating through each combination.

[0180] In an embodiment, one or more combinations of threads and completion time lists may include one combination. In this case, each of the multiple high-efficiency virtual resources includes one of the single virtual machine and virtual machine combinations determined from a list of multiple maximum-efficiency single virtual machines, and each of the multiple high-efficiency virtual resources has a number of cores equal to or greater than the total number of cores that can run a batch job within the maximum completion time. The resource estimation system 100 may also determine a set of virtual machines including the multiple high-efficiency virtual resources, store the multiple high-efficiency virtual resources in a high-efficiency virtual resource list in descending order of efficiency, and select the optimal virtual machine or optimal virtual machine combination from the high-efficiency virtual resource list as the maximum-efficiency virtual resource.

[0181] In an embodiment, one or more combinations in the thread and completion time list may include multiple combinations. In this case, the resource estimation system 100 may, for each of the multiple combinations, (i) iteratively determine a set of virtual machines that include the multiple high-efficiency virtual resources, so that each of the multiple high-efficiency virtual resources includes one of the single virtual machines and virtual machine combinations determined from a list of multiple maximum-efficiency single virtual machines, and each of the multiple high-efficiency virtual resources has a number of cores equal to or greater than the total number of cores that can run a batch job within the maximum completion time. The resource estimation system 100 may, for each of the multiple combinations, select the optimal virtual machine or optimal virtual machine combination by iteratively performing: (ii) storing the multiple high-efficiency virtual resources in a high-efficiency virtual resource list in descending order of efficiency; and (iii) selecting the maximum-efficiency virtual resource as the optimal virtual machine or optimal virtual machine combination based on the multiple high-efficiency virtual resource lists obtained based on the execution of multiple iterations.

[0182] The resource estimation system 100 may also select the optimal virtual machine or optimal virtual machine combination as the most efficient virtual resource by determining the lowest cost virtual machine or virtual machine combination from the list of highly efficient virtual resources.

[0183] Figure 7B shows processing according to various embodiments. For example, the processing 720 shown in Figure 7B may be executed by the resource estimation system 100, for example, by some or all of the model generation subsystem 102, the thread prediction subsystem 103, and the virtual machine selection subsystem 104.

[0184] The process 720 shown in Figure 7B may be implemented by the use of software (e.g., code, instructions, programs), hardware, or a combination thereof, executed by one or more processing units (e.g., processors, cores) of each subsystem. The software may be stored in a non-temporary storage medium (e.g., a memory device). The methods presented in Figure 7B and described below are illustrative and not intended to be limiting. Figure 7B shows various processing operations occurring in a particular order or sequence, but not intended to be limiting. In an alternative embodiment, the process 720 may be executed in some different order, or at least some of the operations may be executed in parallel.

[0185] Continuing with reference to Figure 7B, in operation 722, the resource estimation system 100 may obtain a list of threads and completion times for processing the batch job within the maximum completion time, the list of threads and completion times comprising one or more combinations, each of which comprises a thread estimate and an actual completion time for processing the total number of records of the batch job using the thread estimate.

[0186] In operation 724, the resource estimation system 100 may select an optimal virtual machine or optimal virtual machine combination from the list of available cloud virtual machines using a list of threads and completion times, the optimal virtual machine or optimal virtual machine combination being able to execute batch jobs within the maximum completion time.

[0187] Exemplary Cloud Implementation As mentioned above, Infrastructure as a Service (IaaS) is a specific type of cloud computing. IaaS can be configured to provide virtualized computing resources over a public network (e.g., the internet). In the IaaS model, a cloud computing provider can host infrastructure components (e.g., servers, storage devices, network nodes (e.g., hardware), deployment software, platform virtualization (e.g., hypervisor layer), etc.). Additionally, the IaaS provider may also supply a variety of services associated with these infrastructure components (exemplary services include billing software, monitoring software, logging software, load balancing software, and clustering software, etc.). Since these services can be policy-driven, IaaS users can maintain application availability and performance by implementing policies that promote load balancing.

[0188] In some cases, IaaS customers can access resources and services via a wide area network (WAN), such as the internet, and install other elements of their application stack using the cloud provider's services. For example, a user can log in to the IaaS platform and create virtual machines (VMs), install operating systems (OS) on each virtual machine, deploy middleware such as databases, create storage buckets for workloads and backups, and even install enterprise software on those virtual machines. The customer can then use the provider's services to perform various functions, including distributing network traffic, troubleshooting application issues, monitoring performance, and managing disaster recovery.

[0189] In most cases, the cloud computing model requires the participation of a cloud provider. This cloud provider may, but does not have to be, a third-party service specializing in providing IaaS (e.g., granting, leasing, or selling). Alternatively, an entity could deploy a private cloud and become its own infrastructure service provider.

[0190] In some cases, IaaS deployment is the process of placing a new application or a new version of an application on a prepared application server, etc. It may also include the process of preparing the server (e.g., installing libraries, daemons, etc.). This is often managed by the cloud provider under the hypervisor layer (e.g., servers, storage, network hardware, and virtualization). Therefore, the customer may be responsible for handling (OS), middleware, and / or application deployment (e.g., on self-service virtual machines that can be spun up on demand).

[0191] In some examples, IaaS provisioning represents acquiring the computers or virtual hosts to be used, and may even represent installing the necessary libraries or services on them. In most cases, deployment does not include provisioning, and provisioning may be required to be performed first.

[0192] In some cases, IaaS provisioning presents two distinct challenges. Firstly, there is the initial challenge of provisioning the initial set of infrastructure before anything is operational. Secondly, there is the challenge of developing the existing infrastructure after everything has been provisioned (e.g., adding new services, modifying services, removing services, etc.). In some cases, these two challenges can be addressed by allowing the infrastructure configuration to be defined declaratively. In other words, the infrastructure (e.g., the required components and the way those components interact) can be defined by one or more configuration files. Thus, the overall topology of the infrastructure (e.g., resource dependencies and how each resource works together) can be described declaratively. In some cases, once the topology is defined, a workflow can be generated to form and / or manage the various components described in the configuration files.

[0193] In some cases, infrastructure can consist of many interconnected elements. For example, there may be one or more virtual private clouds (VPCs), also known as core networks (e.g., potential on-demand pools of configurable and / or shared computing resources). In some cases, there may also be one or more inbound / outbound traffic group rules that define how inbound and / or outbound network traffic is configured, as well as provisioning that defines one or more virtual machines (VMs). Other infrastructure elements such as load balancers and databases may also be provisioned. As more and more infrastructure elements are requested and / or added, the infrastructure can evolve little by little.

[0194] In some cases, sequential deployment techniques may be employed to enable the deployment of infrastructure code across various virtual computing environments. Furthermore, the techniques described may enable infrastructure management within these environments. In some examples, a service team may write code that is desirable to be deployed to one or more, but often numerous, different generation environments (e.g., various geographically different locations, sometimes even worldwide). However, in some examples, the infrastructure to which the code is deployed must be configured first. In some cases, provisioning may be performed manually, provisioning tools may be used for provisioning resources, and / or deployment tools may be used to deploy the code after the infrastructure has been provisioned.

[0195] Figure 8 is a block diagram 800 showing an exemplary pattern of an IaaS architecture according to at least one embodiment. A service operator 802 may be communicated with a secure host tenancy 804 which may include a virtual cloud network (VCN) 806 and a secure host subnet 808. In some examples, the service operator 802 may use one or more client computing devices, which may be portable handheld devices (e.g., iPhone®, mobile phones, iPad®, computing tablets, personal digital assistants (PDAs)) or wearable devices (e.g., Google Glass® head-mounted displays) that run software such as Microsoft Windows Mobile® and / or a variety of mobile operating systems such as iOS®, Windows Phone, Android, BlackBerry 8, Palm OS, and can utilize the Internet, email, short message service (SMS), BlackBerry®, or other communication protocols. Alternatively, a client computing device could be a general-purpose personal computer, including personal computers and / or laptop computers running various versions of the Microsoft Windows®, Apple Macintosh®, and / or Linux® operating systems. A client computing device could also be a workstation computer running any of the various commercially available UNIX® or UNIX-like operating systems, including, but not limited to, a variety of GNU / Linux operating systems such as Google Chrome OS.Alternatively or in addition, the client computing device may be any other electronic device, such as a thin client computer, an internet-enabled game system (e.g., a Microsoft Xbox game console with or without a Kinect® gesture input device), and / or a personal messaging device, that can communicate via a network that can access the VCN806 and / or the Internet.

[0196] VCN806 may comprise an LPG810 that can communicate with SSH VCN812 via a local peering gateway (LPG)810 included in Secure Shell (SSH) VCN812. SSH VCN812 may comprise an SSH subnet 814, and SSH VCN812 may communicate with control plane VCN816 via an LPG810 included in control plane VCN816. Furthermore, SSH VCN812 may communicate with data plane VCN818 via LPG810. Control plane VCN816 and data plane VCN818 may be included in a service tenancy 819 owned and / or operated by an IaaS provider.

[0197] The control plane VCN 816 may comprise a control plane demilitarized zone (DMZ) layer 820 that functions as a perimeter network (e.g., part of the corporate network between the corporate intranet and the external network). DMZ-based servers can be limited in responsibility and help deter breaches. The DMZ layer 820 may also comprise a control plane application layer 824 that may include one or more load balancer (LB) subnets 822, application subnets 826, and a control plane data layer 828 that may include database (DB) subnets 830 (e.g., a front-end DB subnet and / or a back-end DB subnet). The LB subnets 822 included in the control plane DMZ layer 820 may be connected to application subnets 826 included in the control plane application layer 824 and an internet gateway 834 that may be included in the control plane VCN 816, while the application subnets 826 may be connected to DB subnets 830, a service gateway 836, and a network address translation (NAT) gateway 838 included in the control plane data layer 828. The control plane VCN816 may include a service gateway 836 and a NAT gateway 838.

[0198] The control plane VCN 816 may comprise a data plane mirror application layer 840 that may include an application subnet 826. The application subnet 826 included in the data plane mirror application layer 840 may comprise a virtual network interface controller (VNIC) 842 that can run a compute instance 844. The compute instance 844 may communicate-couple the application subnet 826 of the data plane mirror application layer 840 to the application subnet 826 that may be included in the data plane application layer 846.

[0199] The data plane VCN818 may comprise a data plane application layer 846, a data plane DMZ layer 848, and a data plane data layer 850. The data plane DMZ layer 848 may comprise an LB subnet 822 that can be connected to the application subnet 826 of the data plane application layer 846 and the internet gateway 834 of the data plane VCN818. The application subnet 826 may be connected to the service gateway 836 and the NAT gateway 838 of the data plane VCN818. The data plane data layer 850 may comprise a DB subnet 830 that can be connected to the application subnet 826 of the data plane application layer 846.

[0200] The Internet gateway 834 of the control plane VCN816 and data plane VCN818 can be connected to a metadata management service 852, which can be connected to the public internet 854. The public internet 854 can be connected to the NAT gateway 838 of the control plane VCN816 and data plane VCN818. The service gateway 836 of the control plane VCN816 and data plane VCN818 can be connected to a cloud service 856.

[0201] In some cases, the service gateway 836 of the control plane VCN816 or data plane VCN818 can make application programming interface (API) calls to the cloud service 856 without going through the public internet 854. API calls from the service gateway 836 to the cloud service 856 can be unidirectional. The service gateway 836 can make API calls to the cloud service 856, and the cloud service 856 can send the requested data to the service gateway 836. However, the cloud service 856 does not need to initiate the API call to the service gateway 836.

[0202] In some examples, secure host tenancy 804 may be directly connected to service tenancy 819, or otherwise isolated. Secure host subnet 808 can communicate with SSH subnet 814 via LPG 810, which can enable bidirectional communication through systems that would otherwise be isolated. Connecting secure host subnet 808 to SSH subnet 814 allows secure host subnet 808 to access other entities within service tenancy 819.

[0203] The control plane VCN816 may enable users of the service tenancy 819 to configure or provision desired resources. Desired resources provisioned in the control plane VCN816 may be deployed or used in the data plane VCN818. In some examples, the control plane VCN816 can be isolated from the data plane VCN818, and the data plane mirror application layer 840 of the control plane VCN816 can communicate with the data plane application layer 846 of the data plane VCN818 via a VNIC 842 which may be included in the data plane mirror application layer 840 and the data plane application layer 846.

[0204] In some examples, a system user, i.e., a customer, can make requests, such as create, read, update, or delete (CRUD) operations, through the public internet 854, and the public internet 854 can send such requests to the metadata management service 852. The metadata management service 852 can send the requests to the control plane VCN 816 through the internet gateway 834. The requests may be received by the LB subnet 822 included in the control plane DMZ layer 820. The LB subnet 822 may determine that the request is valid, and in response to this determination, the LB subnet 822 may send the request to the application subnet 826 included in the control plane application layer 824. If the request is validated and a call to the public internet 854 is required, the call to the public internet 854 may be sent to a NAT gateway 838 that can make a call to the public internet 854. Metadata that is deemed desirable to be stored by the request may be stored in the DB subnet 830.

[0205] In some examples, the data plane mirror application layer 840 can facilitate direct communication between the control plane VCN816 and the data plane VCN818. For example, it may be desirable that configuration changes, updates, or other preferred modifications be applied to resources contained in the data plane VCN818. Through VNIC842, the control plane VCN816 can perform configuration changes, updates, or other preferred modifications to resources by communicating directly with the resources contained in the data plane VCN818.

[0206] In some embodiments, the control plane VCN816 and the data plane VCN818 may be included in the service tenancy 819. In this case, the system user, i.e., the customer, does not have to own either the control plane VCN816 or the data plane VCN818, or does not have either of them running. Alternatively, the IaaS provider may own both the control plane VCN816 and the data plane VCN818, or have both running, and both may be included in the service tenancy 819. This embodiment may enable network isolation that can prevent interaction between the user, i.e., the customer, and other users, i.e., other customers' resources. Furthermore, this embodiment may enable private storage of databases by the system user, i.e., the customer, without having to rely on the public internet 854, which may not have the desired level of threat prevention for storage.

[0207] In another embodiment, the LB subnet 822 included in the control plane VCN 816 can be configured to receive signals from the service gateway 836. In this embodiment, the control plane VCN 816 and the data plane VCN 818 may be configured to be invoked by the IaaS provider's customer without calling the public internet 854. The IaaS provider's customer is likely to prefer this embodiment because the database used by the customer can be stored in a service tenancy 819 that is controlled by the IaaS provider and can be isolated from the public internet 854.

[0208] Figure 9 is a block diagram 900 showing another exemplary pattern of an IaaS architecture according to at least one embodiment. A service operator 902 (e.g., service operator 802 in Figure 8) may be connected to a secure host tenancy 904 (e.g., secure host tenancy 804 in Figure 8), which may include a virtual cloud network (VCN) 906 (e.g., VCN806 in Figure 8) and a secure host subnet 908 (e.g., secure host subnet 808 in Figure 8). The VCN 906 may comprise an LPG 910 that can be connected to the SSH VCN 912 via a local peering gateway (LPG) 810 (e.g., LPG810 in Figure 8), which is included in the Secure Shell (SSH) VCN 912 (e.g., SSH VCN812 in Figure 8). SSH VCN912 may comprise an SSH subnet 914 (e.g., SSH subnet 814 in Figure 8), and SSH VCN912 may be communicated with control plane VCN916 via an LPG 910 included in control plane VCN916 (e.g., control plane VCN816 in Figure 8). Control plane VCN916 may be included in service tenancy 919 (e.g., service tenancy 819 in Figure 8), and data plane VCN918 (e.g., data plane VCN818 in Figure 8) may be included in customer tenancy 921, which may be owned or operated by the system's users, i.e., customers.

[0209] The control plane VCN916 may comprise a control plane DMZ layer 920 (e.g., control plane DMZ layer 820 in Figure 8) which may include an LB subnet 922 (e.g., LB subnet 822 in Figure 8), a control plane application layer 924 (e.g., control plane application layer 824 in Figure 8) which may include an application subnet 926 (e.g., application subnet 826 in Figure 8), and a control plane data layer 928 (e.g., control plane data layer 828 in Figure 8) which may include a DB subnet 930 (e.g., similar to the database (DB) subnet 830 in Figure 8). The LB subnet 922 included in the control plane DMZ layer 920 is connected to the application subnet 926 included in the control plane application layer 924 and the Internet gateway 934 (e.g., Internet gateway 834 in Figure 8) which may be included in the control plane VCN 916. The application subnet 926 may be connected to the DB subnet 930, the service gateway 936 (e.g., service gateway 836 in Figure 8), and the Network Address Translation (NAT) gateway 938 (e.g., NAT gateway 838 in Figure 8) included in the control plane data layer 928. The control plane VCN 916 may include the service gateway 936 and the NAT gateway 938.

[0210] The control plane VCN916 may comprise a data plane mirror application layer 940 (e.g., the data plane mirror application layer 840 in Figure 8) which may include an application subnet 926. The application subnet 926 included in the data plane mirror application layer 940 may comprise a virtual network interface controller (VNIC) 942 (e.g., VNIC842) capable of running a compute instance 944 (e.g., similar to the compute instance 844 in Figure 8). The compute instance 944 may facilitate communication between the application subnet 926 of the data plane mirror application layer 940 and the application subnet 926 that may be included in the data plane application layer 946 via the VNIC942 included in the data plane mirror application layer 940 and the VNIC942 included in the data plane application layer 946 (e.g., the data plane application layer 846 in Figure 8).

[0211] The Internet gateway 934 included in the control plane VCN916 can be connected to a metadata management service 952 (e.g., metadata management service 852 in Figure 8), which can be connected to the public internet 954 (e.g., public internet 854 in Figure 8). The public internet 954 can be connected to a NAT gateway 938 included in the control plane VCN916. The service gateway 936 included in the control plane VCN916 can be connected to a cloud service 956 (e.g., cloud service 856 in Figure 8).

[0212] In some examples, the data plane VCN918 may be included in a customer tenancy 921. In this case, the IaaS provider may provide a control plane VCN916 for each customer, or the IaaS provider may configure a unique compute instance 944 included in a service tenancy 919 for each customer. Each compute instance 944 may enable communication between the control plane VCN916 included in the service tenancy 919 and the data plane VCN918 included in the customer tenancy 921. The compute instance 944 enables resources provisioned in the control plane VCN916 included in the service tenancy 919 to be deployed or used in the data plane VCN918 included in the customer tenancy 921.

[0213] In another example, the IaaS provider's customer may have a database located in customer tenancy 921. In this example, the control plane VCN916 may comprise a data plane mirror application layer 940, which may include application subnet 926. The data plane mirror application layer 940 may reside in data plane VCN918, but does not have to reside in data plane VCN918. That is, the data plane mirror application layer 940 may be accessible to customer tenancy 921, but does not have to reside in data plane VCN918, nor is it owned or operated by the IaaS provider's customer. The data plane mirror application layer 940 may be configured to make calls to data plane VCN918, but may be configured not to make calls to any entities included in control plane VCN916. The customer is expected to want to deploy or use resources in the data plane VCN918 that are provisioned in the control plane VCN916, and the data plane mirror application layer 940 can facilitate the deployment or other use of the resources desired by the customer.

[0214] In some embodiments, a customer of the IaaS provider can apply filters to the data plane VCN918. In this embodiment, the customer can determine what the data plane VCN918 can access, and may also restrict access from the data plane VCN918 to the public internet 954. The IaaS provider may not be able to apply filters or control the data plane VCN918's access to any external network or database. The application of filters and controls by the customer to the data plane VCN918 included in the customer tenancy 921 may help isolate the data plane VCN918 from other customers and the public internet 954.

[0215] In some embodiments, the cloud service 956 becomes accessible by a call from the service gateway 936 to a service that could not exist on the public internet 954, the control plane VCN916, or the data plane VCN918. The connection between the cloud service 956 and the control plane VCN916 or data plane VCN918 does not have to be live or continuous. The cloud service 956 may reside on different networks owned or operated by the IaaS provider. The cloud service 956 may be configured to receive calls from the service gateway 936 and may be configured not to receive calls from the public internet 954. Some cloud services 956 may be isolated from other cloud services 956, and the control plane VCN916 may be isolated from cloud services 956 that could not be in the same region as the control plane VCN916. For example, the control plane VCN916 may be located in "Region 1", and the cloud service "Deployment 8" may be located in Region 1 and "Region 2". If a call to deployment 8 is made by a service gateway 936 included in the control plane VCN916 located in region 1, the call may be sent to deployment 8 in region 1. In this example, the control plane VCN916 or deployment 8 in region 1 does not need to be communicated with or in communication with deployment 8 in region 2.

[0216] Figure 10 is a block diagram 1000 showing another exemplary pattern of an IaaS architecture according to at least one embodiment. A service operator 1002 (e.g., service operator 802 in Figure 8) may be connected to a secure host tenancy 1004 (e.g., secure host tenancy 804 in Figure 8), which may include a virtual cloud network (VCN) 1006 (e.g., VCN806 in Figure 8) and a secure host subnet 1008 (e.g., secure host subnet 808 in Figure 8). VCN 1006 may comprise an LPG 1010 (e.g., LPG810 in Figure 8) which is included in SSH VCN 1012 (e.g., SSH VCN812 in Figure 8). SSH VCN1012 may comprise an SSH subnet 1014 (for example, SSH subnet 814 in Figure 8), and SSH VCN1012 may be connected to control plane VCN1016 via an LPG 1010 included in control plane VCN1016 (for example, control plane VCN816 in Figure 8), and may also be connected to data plane VCN1018 via an LPG 1010 included in data plane VCN1018 (for example, data plane 818 in Figure 8). Control plane VCN1016 and data plane VCN1018 may be included in service tenancy 1019 (for example, service tenancy 819 in Figure 8).

[0217] The control plane VCN1016 may comprise a control plane DMZ layer 1020 (e.g., control plane DMZ layer 820 in Figure 8) which may include a load balancer (LB) subnet 1022 (e.g., LB subnet 822 in Figure 8), a control plane application layer 1024 (e.g., control plane application layer 824 in Figure 8) which may include an application subnet 1026 (e.g., similar to application subnet 826 in Figure 8), and a control plane data layer 1028 (e.g., control plane data layer 828 in Figure 8) which may include a DB subnet 1030. The LB subnet 1022 included in the control plane DMZ layer 1020 is connected to the application subnet 1026 included in the control plane application layer 1024 and the Internet gateway 1034 (e.g., Internet gateway 834 in Figure 8) which may be included in the control plane VCN 1016. The application subnet 1026 may be connected to the DB subnet 1030 included in the control plane data layer 1028, the service gateway 1036 (e.g., the service gateway in Figure 8), and the network address translation (NAT) gateway 1038 (e.g., NAT gateway 838 in Figure 8). The control plane VCN 1016 may include the service gateway 1036 and the NAT gateway 1038.

[0218] The data plane VCN 1018 may comprise a data plane application layer 1046 (e.g., data plane application layer 846 in Figure 8), a data plane DMZ layer 1048 (e.g., data plane DMZ layer 848 in Figure 8), and a data plane data layer 1050 (e.g., data plane data layer 850 in Figure 8). The data plane DMZ layer 1048 may comprise an LB subnet 1022 that can be connected to the trusted application subnet 1060 and the untrusted application subnet 1062 of the data plane application layer 1046, as well as the internet gateway 1034 included in the data plane VCN 1018. The trusted application subnet 1060 may be connected to the service gateway 1036 included in the data plane VCN 1018, the NAT gateway 1038 included in the data plane VCN 1018, and the DB subnet 1030 included in the data plane data layer 1050. The non-trusted application subnet 1062 can be connected to the service gateway 1036 included in the data plane VCN 1018 and the DB subnet 1030 included in the data plane data layer 1050. The data plane data layer 1050 may include the DB subnet 1030, which can be connected to the service gateway 1036 included in the data plane VCN 1018.

[0219] The untrusted application subnet 1062 may comprise one or more primary VNICs 1064(1)-(N) that can be connected to tenant virtual machines (VMs) 1066(1)-(N). Each tenant virtual machine 1066(1)-(N) may comprise an application subnet 1067(1)-(N) that may comprise each container output VCN 1068(1)-(N) that may comprise each customer tenancy 1070(1)-(N). The secondary VNICs 1072(1)-(N) may facilitate communication between the untrusted application subnet 1062, which is included in the data plane VCN 1018, and the application subnets included in the container output VCN 1068(1)-(N). Each container output VCN 1068(1)-(N) may comprise a NAT gateway 1038 that can be connected to the public internet 1054 (e.g., public internet 854 in Figure 8).

[0220] The Internet gateway 1034 included in the control plane VCN1016 and data plane VCN1018 can be connected to a metadata management service 1052 (e.g., metadata management service 852 in Figure 8), which can be connected to the public internet 1054. The public internet 1054 can be connected to a NAT gateway 1038 included in the control plane VCN1016 and data plane VCN1018. The service gateway 1036 included in the control plane VCN1016 and data plane VCN1018 can be connected to a cloud service 1056.

[0221] In some embodiments, the data plane VCN 1018 may be integrated with a customer tenancy 1070. This integration may be useful or desirable for the IaaS provider's customer, for example, if they may want support for executing code. The customer may provide code that could be disruptive to execute, communicate with other customer resources, or have undesirable effects. In response, the IaaS provider may determine whether to execute the code provided to the IaaS provider by the customer.

[0222] In some examples, an IaaS provider's customer may allow temporary network access to the IaaS provider and request functionality granted to the dataplane application layer 1046. The code that performs this functionality may run in a virtual machine 1066(1)-(N), and such code may not be configured to run elsewhere on the dataplane VCN 1018. Each virtual machine 1066(1)-(N) may be connected to a single customer tenancy 1070. Each container 1071(1)-(N) contained within a virtual machine 1066(1)-(N) may be configured to run code. In this case, a double isolation may exist (for example, the container 1071(1)-(N) that runs the code may be contained within a virtual machine 1066(1)-(N) that is at least in a non-trusted application subnet 1062), which may help prevent damage to the IaaS provider's network or the networks of different customers by erroneous or undesirable code. Container 1071(1)-(N) may be communication-coupled to customer tenancy 1070, and may be configured to send or receive data to or from customer tenancy 1070. Container 1071(1)-(N) may be configured not to send or receive data to or from any other entity in the data plane VCN 1018. Upon completion of code execution, the IaaS provider may disable or discard container 1071(1)-(N).

[0223] In some embodiments, the trusted application subnet 1060 may be configured to execute code owned or operated by the IaaS provider. In this embodiment, the trusted application subnet 1060 may be connected to the DB subnet 1030 and may be configured to perform CRUD operations in the DB subnet 1030. The untrusted application subnet 1062 may be connected to the DB subnet 1030, but in this embodiment, the untrusted application subnet 1062 may be configured to perform read operations in the DB subnet 1030. Containers 1071(1)-(N) contained in each customer's virtual machine 1066(1)-(N) and capable of executing code from the customer may not be connected to the DB subnet 1030.

[0224] In other embodiments, the control plane VCN1016 and the data plane VCN1018 do not have to be directly connected. In this embodiment, direct communication between the control plane VCN1016 and the data plane VCN1018 is not required. However, communication can be performed indirectly by at least one method. An LPG1010 that can facilitate communication between the control plane VCN1016 and the data plane VCN1018 may be established by the IaaS provider. In another example, the control plane VCN1016 or the data plane VCN1018 can make a call to the cloud service 1056 via the service gateway 1036. For example, a call from the control plane VCN1016 to the cloud service 1056 may include a request for a service that can communicate with the data plane VCN1018.

[0225] Figure 11 is a block diagram 1100 showing another exemplary pattern of an IaaS architecture according to at least one embodiment. A service operator 1102 (e.g., service operator 802 in Figure 8) may be connected to a secure host tenancy 1104 (e.g., secure host tenancy 804 in Figure 8), which may include a virtual cloud network (VCN) 1106 (e.g., VCN806 in Figure 8) and a secure host subnet 1108 (e.g., secure host subnet 808 in Figure 8). VCN 1106 may comprise an LPG 1110 (e.g., LPG810 in Figure 8) which is included in SSH VCN 1112 (e.g., SSH VCN812 in Figure 8). SSH VCN1112 may comprise SSH subnet 1114 (for example, SSH subnet 814 in Figure 8), and SSH VCN1112 may be connected to control plane VCN1116 via LPG1110 included in control plane VCN1116 (for example, control plane VCN816 in Figure 8), and may also be connected to data plane VCN1118 via LPG1110 included in data plane VCN1118 (for example, data plane 818 in Figure 8). Control plane VCN1116 and data plane VCN1118 may be included in service tenancy 1119 (for example, service tenancy 819 in Figure 8).

[0226] The control plane VCN1116 may comprise a control plane DMZ layer 1120 (e.g., control plane DMZ layer 820 in Figure 8) which may include an LB subnet 1122 (e.g., LB subnet 822 in Figure 8), a control plane application layer 1124 (e.g., control plane application layer 824 in Figure 8) which may include an application subnet 1126 (e.g., application subnet 826 in Figure 8), and a control plane data layer 1128 (e.g., control plane data layer 828 in Figure 8) which may include a DB subnet 1130 (e.g., DB subnet 1030 in Figure 10). The LB subnet 1122 included in the control plane DMZ layer 1120 is connected to the application subnet 1126 included in the control plane application layer 1124 and the Internet gateway 1134 (e.g., Internet gateway 834 in Figure 8) which may be included in the control plane VCN 1116. The application subnet 1126 may be connected to the DB subnet 1130, the service gateway 1136 (e.g., the service gateway in Figure 8), and the Network Address Translation (NAT) gateway 1138 (e.g., NAT gateway 838 in Figure 8) included in the control plane data layer 1128. The control plane VCN 1116 may comprise the service gateway 1136 and the NAT gateway 1138.

[0227] The data plane VCN 1118 may comprise a data plane application layer 1146 (e.g., data plane application layer 846 in Figure 8), a data plane DMZ layer 1148 (e.g., data plane DMZ layer 848 in Figure 8), and a data plane data layer 1150 (e.g., data plane data layer 850 in Figure 8). The data plane DMZ layer 1148 may comprise a trusted application subnet 1160 (e.g., trusted application subnet 1060 in Figure 10) and an untrusted application subnet 1162 (e.g., untrusted application subnet 1062 in Figure 10) of the data plane application layer 1146, as well as an LB subnet 1122 that can be connected to an internet gateway 1134 included in the data plane VCN 1118. Trusted application subnet 1160 can be connected to service gateway 1136 included in data plane VCN 1118, NAT gateway 1138 included in data plane VCN 1118, and DB subnet 1130 included in data plane data layer 1150. Non-trusted application subnet 1162 can be connected to service gateway 1136 included in data plane VCN 1118 and DB subnet 1130 included in data plane data layer 1150. Data plane data layer 1150 may include DB subnet 1130 that can be connected to service gateway 1136 included in data plane VCN 1118.

[0228] The untrusted application subnet 1162 may comprise primary VNICs 1164(1)-(N) that can be connected to tenant virtual machines (VMs) 1166(1)-(N) residing within the untrusted application subnet 1162. Each tenant virtual machine 1166(1)-(N) can execute code in each of its containers 1167(1)-(N) and can be connected to an application subnet 1126 that may be included in the data plane application layer 1146 that may be included in the container output VCN 1168. The secondary VNICs 1172(1)-(N) can facilitate communication between the untrusted application subnet 1162 included in the data plane VCN 1118 and the application subnet included in the container output VCN 1168. The container output VCN may comprise a NAT gateway 1138 that can be connected to the public internet 1154 (e.g., the public internet 854 in Figure 8).

[0229] The Internet gateway 1134 included in the control plane VCN1116 and data plane VCN1118 can be connected to a metadata management service 1152 (e.g., metadata management service 852 in Figure 8), which can be connected to the public internet 1154. The public internet 1154 can be connected to a NAT gateway 1138 included in the control plane VCN1116 and data plane VCN1118. The service gateway 1136 included in the control plane VCN1116 and data plane VCN1118 can be connected to a cloud service 1156.

[0230] In some examples, the pattern shown by the architecture of block diagram 1100 in Figure 11 is considered an exception to the pattern shown by the architecture of block diagram 1000 in Figure 10, and is considered desirable for the IaaS provider's customers when the IaaS provider cannot communicate directly with the customers (e.g., in an unconnected region). Each container 1167(1)-(N) contained within a virtual machine 1166(1)-(N) for each customer is accessible to the customer in real time. Each container 1167(1)-(N) may be configured to make calls to each of the secondary VNICs 1172(1)-(N) contained in the application subnet 1126 of the data plane application layer 1146, which may be contained in the container output VCN 1168. The secondary VNICs 1172(1)-(N) can send calls to the NAT gateway 1138, which may send calls to the public internet 1154. In this example, the containers 1167(1)-(N), which are accessible to customers in real time, can be isolated from the control plane VCN1116, as well as from other entities included in the data plane VCN1118. Furthermore, the containers 1167(1)-(N) can also be isolated from other customers' resources.

[0231] In another example, a customer can use container 1167(1)-(N) to invoke cloud service 1156. In this example, the customer can execute code in container 1167(1)-(N) that requests a service from cloud service 1156. Container 1167(1)-(N) can send this request to secondary VNIC 1172(1)-(N), secondary VNIC 1172(1)-(N) can send the request to the NAT gateway, and the NAT gateway can send the request to the public internet 1154. The public internet 1154 can send the request to LB subnet 1122, which is included in control plane VCN 1116, via internet gateway 1134. In response to the determination that the request is valid, the LB subnet can send the request to application subnet 1126, and application subnet 1126 can send the request to cloud service 1156 via service gateway 1136.

[0232] It should be noted that the IaaS architectures 800, 900, 1000, and 1100 shown in the drawings may have components other than those shown. Furthermore, the embodiments shown in the drawings are only some examples of cloud infrastructure systems that may encompass one embodiment of this disclosure. In some other embodiments, the IaaS system may have more or fewer components than shown, may combine two or more components, or may have components in different configurations or arrangements.

[0233] In one embodiment, the IaaS system described herein may include a set of applications, middleware, and database service offerings delivered to customers in a self-service, subscription-based, elastically scalable, reliable, highly available, and secure manner. An example of such an IaaS system is Oracle Cloud Infrastructure (OCI) provided by the Assignee.

[0234] Figure 12 shows an exemplary computer system 1200 in which various embodiments can be realized. The computer system 1200 may be used to realize any of the computer systems described above. As shown in the drawing, the computer system 1200 comprises a processing unit 1204 that communicates with a number of peripheral subsystems via a bus subsystem 1202. The peripheral subsystems may include a processing acceleration unit 1206, an I / O subsystem 1208, a storage subsystem 1218, and a communication subsystem 1224. The storage subsystem 1218 comprises a tangible computer-readable storage medium 1222 and a system memory 1210.

[0235] The bus subsystem 1202 provides a mechanism for various components and subsystems of the computer system 1200 to communicate with each other as intended. Although the bus subsystem 1202 is schematically shown as a single bus, alternative embodiments of the bus subsystem may utilize multiple buses. The bus subsystem 1202 may be one of several types of bus structures, including a memory bus or memory controller, a peripheral bus, and a local bus, using any of a variety of bus architectures. For example, such architectures may include industry standard architecture (ISA) buses, microchannel architecture (MCA) buses, extended ISA (EISA) buses, video electronics standards (VESA) local buses, and peripheral interconnect (PCI) buses, which can be implemented as mezzanine buses manufactured according to the IEEE P1386.1 standard.

[0236] A processing unit 1204, which can be implemented as one or more integrated circuits (e.g., conventional microprocessors or microcontrollers), controls the operation of the computer system 1200. One or more processors may be included in the processing unit 1204. These processors may include single-core or multi-core processors. In some embodiments, the processing unit 1204 may be implemented as one or more independent processing units 1232 and / or 1234, each containing a single-core or multi-core processor. In other embodiments, the processing unit 1204 may be implemented as a quad-core processing unit formed by integrating two dual-core processors onto a single chip.

[0237] In various embodiments, the processing unit 1204 can execute a variety of programs in response to program code and maintain multiple concurrently running programs or processes. At any given time, some or all of the program code to be executed may reside in the processor 1204 and / or the storage subsystem 1218. With suitable programming, the processor 1204 can provide the various functions described above. The computer system 1200 may further comprise a processing acceleration unit 1206, which may include a digital signal processor (DSP), a dedicated processor, and / or similar.

[0238] The I / O subsystem 1208 may include user interface input devices and user interface output devices. User interface input devices may include pointing devices such as keyboards, mice or trackballs, touchpads or touchscreens integrated into displays, scroll wheels, click wheels, dials, buttons, switches, keypads, audio input devices with voice command recognition systems, microphones, and other types of input devices. User interface input devices may include motion detection and / or gesture recognition devices such as Microsoft Kinect® motion sensors that enable user control and interaction with input devices such as Microsoft Xbox® 360 game controllers through a natural user interface using gestures and voice commands. User interface input devices may also include eye gesture recognition devices such as Google Glass® blink detectors that detect the user's eye activity (e.g., blinking while taking photos and / or selecting menus) and convert eye gestures into input to an input device (e.g., Google Glass®). Furthermore, the user interface input device may include a voice recognition detection device that enables user interaction with a voice recognition system (e.g., Siri® Navigator) via voice commands.

[0239] Furthermore, user interface input devices may include, but are not limited to, three-dimensional (3D) mice, joysticks or pointing sticks, gamepads and graphic tablets, as well as auditory or visual devices such as speakers, digital cameras, digital video cameras, portable media players, webcams, image scanners, fingerprint scanners, barcode readers, 3D scanners, 3D printers, laser rangefinders, and eye-tracking devices. User interface input devices may also include, for example, medical imaging input devices such as computed tomography, magnetic resonance imaging, positron emission tomography, and medical ultrasound devices. Additionally, user interface input devices may include, for example, audio input devices such as MIDI keyboards and digital musical instruments.

[0240] User interface output devices may include non-visual displays such as display subsystems, indicator lights, or audio output devices. Display subsystems may also include flat panel devices such as cathode ray tubes (CRTs), liquid crystal displays (LCDs), or plasma displays, projection devices, touchscreens, etc. Generally, the use of the term “output device” is intended to include all conceivable types of devices and mechanisms for outputting information from the computer system 1200 to a user or another computer. For example, user interface output devices may include, but are not limited to, a variety of display devices that visually convey text, graphics, and audio / video information, such as monitors, printers, speakers, headphones, car navigation systems, plotters, audio output devices, and modems.

[0241] The computer system 1200 may also include a storage subsystem 1218 containing software elements, which are shown here as being located in the system memory 1210. The system memory 1210 may store program instructions that can be loaded into the processing unit 1204 and executed, as well as data generated during the execution of these programs.

[0242] Depending on the configuration and type of the computer system 1200, the system memory 1210 may be volatile (e.g., random access memory (RAM)) and / or non-volatile (e.g., read-only memory (ROM), flash memory). The RAM is usually immediately accessible to the processing unit 1204 and / or contains data and / or program services currently operated and executed by the processing unit 1204. In some embodiments, the system memory 1210 may include several different types of memory, such as static random access memory (SRAM) or dynamic random access memory (DRAM). In some embodiments, a basic input / output system (BIOS) containing basic routines that help transmit information between elements within the computer system 1200 at startup, etc., is usually stored in ROM. Also, as a non-limiting example, the system memory 1210 may include application programs 1212, program data 1214, and an operating system 1216, which may include client applications, web browsers, middle-tier applications, relational database management systems (RDBMS), etc. As an example, Operating System 1216 may include various versions of Microsoft Windows®, Apple Macintosh®, and / or Linux operating systems, various commercially available UNIX® or UNIX-like operating systems (including, but not limited to, various GNU / Linux operating systems, Google Chrome® OS, etc.), and / or mobile operating systems such as iOS, Windows® Phone, Android® OS, BlackBerry® OS, and Palm® OS.

[0243] Furthermore, the storage subsystem 1218 may provide a tangible computer-readable storage medium for storing basic programming and data constructs that provide the functionality of some embodiments. Software (programs, code services, instructions) that, when executed by the processor, provide the above-described functionality may be stored in the storage subsystem 1218. These software services or instructions may be executed by the processing unit 1204. The storage subsystem 1218 may also provide a repository for storing data used in accordance with this disclosure.

[0244] Furthermore, the storage subsystem 1218 may include a computer-readable storage medium reader 1220 which can be further connected to the computer-readable storage medium 1222. The computer-readable storage medium 1222 may, as a whole, optionally in combination with the system memory 1210, comprehensively represent a storage medium for temporarily and / or permanently containing, storing, transmitting, and retrieving computer-readable information, as well as remote, local, fixed, and / or removable storage devices.

[0245] Furthermore, the computer-readable storage medium 1222 containing code or a portion of code may include, but is not limited to, any suitable medium known or used in the art, including storage and communication media, and volatile and non-volatile, removable and non-removable media realized in any method or technique for storing and / or transmitting information. This may include tangible computer-readable storage media such as RAM, ROM, electronically erasable programmable ROM (EEPROM), flash memory or other memory technologies, CD-ROM, digital multipurpose disk (DVD) or optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices, or other tangible computer-readable media. This may also include intangible computer-readable media such as any other medium that can be used for transmitting data signals, data transmission or desired information and is accessible by the computer system 1200.

[0246] For example, the computer-readable storage medium 1222 may include a hard disk drive that reads and writes to a non-removable non-volatile magnetic medium, a magnetic disk drive that reads and writes to a removable non-volatile magnetic disk, and an optical disk drive that reads and writes to a removable non-volatile optical disk such as a CD-ROM, DVD, Blu-ray® disc, or other optical medium. The computer-readable storage medium 1222 may include, but is not limited to, a Zip® drive, a flash memory card, a Universal Serial Bus (USB) flash drive, a Secure Digital (SD) card, a DVD disc, a digital videotape, etc. Furthermore, the computer-readable storage medium 1222 may include static state drives (SSDs) based on non-volatile memory, such as flash memory-based SSDs, enterprise flash drives, and static state ROMs; SSDs based on volatile memory, such as static state RAM, dynamic RAM, static RAM, DRAM-based SSDs, and magnetoresistive RAM (MRAM) SSDs; and hybrid SSDs using a combination of DRAM and flash memory-based SSDs. The disk drive and its respective associated computer-readable medium may provide non-volatile storage for computer-readable instructions, data structures, program services, and other data to the computer system 1200.

[0247] The communication subsystem 1224 provides an interface to other computer systems and networks. The communication subsystem 1224 functions as an interface for sending and receiving data between the computer system 1200 and other systems. For example, the communication subsystem 1224 may enable the computer system 1200 to connect to one or more devices via the Internet. In some embodiments, the communication subsystem 1224 may comprise a wireless voice and / or data network (using, for example, cellular technology such as 3G, 4G, 5G, or EDGE (Enhanced Data Rates for Global Evolution), advanced data network technology, WiFi (IEEE 802.11 family standard), or other mobile communication technology, or any combination thereof), a Global Positioning System (GPS) receiver component, and / or a radio frequency (RF) transceiver component for accessing other components. In some embodiments, the communication subsystem 1224 may provide a wired network connection (e.g., Ethernet®) as an addition to or alternative to the wireless interface.

[0248] In addition, in some embodiments, the communication subsystem 1224 can receive incoming communications in the form of structured and / or unstructured data feeds 1226, event streams 1228, event updates 1230, etc., on behalf of one or more users who can use the computer system 1200.

[0249] For example, the communication subsystem 1224 may be configured to receive data feeds 1226 in real time from users of social networks, and / or from users of other communication services, such as Twitter® feeds, Facebook® updates, rich site summary (RSS) feeds, and / or real-time updates from one or more third-party sources.

[0250] Furthermore, the communication subsystem 1224 may be configured to receive data in the form of a continuous data stream, which may include an event stream 1228 and / or event update 1230 of real-time events, which may be continuous or virtually infinite with no explicit end. Examples of applications that generate continuous data include, for example, sensor data applications, financial tickers, network performance measurement tools (e.g., network monitoring and traffic management applications), clickstream analysis tools, and automotive traffic monitoring.

[0251] Furthermore, the communication subsystem 1224 may be configured to output structured and / or unstructured data feeds 1226, event streams 1228, event updates 1230, etc., to one or more databases that can communicate with one or more streaming data source computers connected to the computer system 1200.

[0252] The computer system 1200 can be one of a variety of types, including portable handheld devices (e.g., iPhone® mobile phones, iPad® computing tablets, PDAs, etc.), wearable devices (e.g., Google Glass® head-mounted displays), PCs, workstations, mainframes, kiosks, server racks, or any other data processing systems.

[0253] Due to the constantly changing nature of computers and networks, the description of the illustrated computer system 1200 is intended only as an example. Many other configurations with more or fewer components than the illustrated system are possible. For example, the use of customized hardware and / or the implementation of specific elements in hardware, firmware, software (including applets), or combinations thereof are also possible. Furthermore, connections to other computing devices such as network input / output devices may be employed. Based on the disclosures and teachings contained herein, those skilled in the art will recognize other methods and / or ways of realizing various embodiments.

[0254] While specific embodiments have been described above, various improvements, modifications, alternative configurations, and equivalents are also included in the scope of this disclosure. The embodiments are not limited to operation within a few specific data processing environments, but can freely operate within multiple data processing environments. Furthermore, while the embodiments have been described using a specific set of transactions and steps, it will be clear to those skilled in the art that the scope of this disclosure is not limited to the set of transactions and steps described. The various features and aspects of the embodiments described above may be used individually or in combination.

[0255] Furthermore, while embodiments have been described using specific combinations of hardware and software, it will be recognized that other combinations of hardware and software are also included in the scope of this disclosure. Embodiments may be implemented in hardware only, in software only, or by using a combination of these. The various processes described herein may be implemented on the same processor or on any combination of different processors. Thus, where a component or service is described as being configured to perform some operation, such configuration may be achieved, for example, by designing electronic circuits to perform this operation, by programming programmable electronic circuits (such as a microprocessor) to perform this operation, or by any combination thereof. Processes may be communicative using a variety of techniques, including but not limited to conventional techniques for inter-process communication. Also, different pairs of processes may use different techniques, or the same pair of processes may use different techniques at different times.

[0256] Accordingly, this specification and the drawings are intended to be illustrative and not limiting in any way. However, it is clear that additions, reductions, deletions, and other improvements and modifications can be made without departing from the broad idea and scope set forth in the claims. For this reason, specific embodiments of this disclosure have been described, but these are not intended to be limiting in any way. Various improvements and equivalents are included in the following claims.

[0257] The terms "a", "an", and "the", as well as similar referents, in the context of describing the disclosed embodiments (particularly in the context of the following claims) are to be construed as applying to both the singular and plural forms, unless otherwise indicated herein or there is a clear contradiction in the context. The terms "including", "having", "comprising", and "containing" are to be construed as open-ended terms (i.e., meaning "including but not limited to") unless otherwise stated. The term "connected" is to be construed as including, attached to, or integrally coupled with, even if there is something intervening, in part or in whole. The recitation of a range of values herein is intended to function only as a concise way of referring individually to each separate value within the range, and each separate value is incorporated herein as if it were individually recited herein, unless otherwise indicated herein. Unless otherwise indicated or there is a clear contradiction in the context, all methods described herein can be performed in any suitable order. The use of any example or exemplary language (e.g., "such as") in this specification is for illustrative purposes only to facilitate understanding of the embodiments and does not limit the scope of the present disclosure unless otherwise claimed. No representation in this specification is to be construed as indicating that any non-claimed element is essential to the practice of the present disclosure.

[0258] Unless otherwise specified, disjunctive language such as the phrase "at least one of X, Y, or Z" is intended to be understood within the context in which it is commonly used to present that a clause, term, etc. can be any one of X, Y, Z or any combination thereof (e.g., X, Y, and / or Z). Thus, such disjunctive language is generally not intended, and shall not be construed, to imply that some embodiments require the presence of at least one of each of at least one of X, at least one of Y, or at least one of Z.

[0259] This specification describes preferred embodiments of the present disclosure, including the best mode known as of the filing date of this disclosure. Persons having ordinary skill in the art will recognize, upon reading the above description, that variations of these preferred embodiments may become apparent. Persons having ordinary skill in the art may adopt such variations as appropriate, and the present disclosure may be practiced otherwise than as specifically described herein. Accordingly, the present disclosure includes all improvements and equivalents of the subject matter recited in the claims appended hereto as permitted by applicable law. Further, unless otherwise indicated herein, any combination of the possible variations of the above-described elements is included in the present disclosure.

[0260] All references, including publications, patent applications, and patents cited herein, are incorporated by reference herein to the same extent as if each reference were individually and specifically indicated to be incorporated by reference and were set forth in its entirety herein.

[0261] While the above specification describes aspects of the disclosure with reference to specific embodiments, those skilled in the art will recognize that the disclosure is not limited thereto. The various features and aspects of the disclosure described above may be used individually or in combination. Furthermore, embodiments may be used in any number of environments and applications beyond those described herein, without departing from the broader concept and scope of this specification. Accordingly, this specification and the drawings should be considered illustrative and not limiting.

Claims

1. A computer execution method, This includes estimating the number of threads that will execute a batch job within the maximum completion time by executing a thread estimation algorithm, and executing the thread estimation algorithm is This involves using a machine learning (ML) model to determine one or more combinations of sets, each of which includes a thread estimate and an actual completion time for processing the total number of records in the batch job using the thread estimate, wherein the actual completion time is less than or equal to the maximum completion time. The method further includes storing the set, which includes one or more of the aforementioned combinations, in a thread and completion time list. A computer execution method wherein the thread and completion time list is used to select the optimal virtual machine or optimal virtual machine combination based on the list of available cloud virtual machines, and the optimal virtual machine or optimal virtual machine combination is capable of executing the batch job within the maximum completion time.

2. Executing the aforementioned thread estimation algorithm means From 2, for each number of threads among the multiple number of threads of the predetermined maximum number of threads, Based on the current number of threads in the current iteration among multiple iterations, the thread initialization time for processing the batch job is determined, The process execution time is determined using the aforementioned maximum completion time and thread initialization time. The process execution time is given as input to the ML model, Based on the execution time of the input process, the number of predicted records to be processed by the current number of threads is obtained as the output of the ML model, The process involves determining whether the predicted number of records is greater than or equal to the expected number of records that can be processed using the current number of threads, If the predicted number of records is greater than or equal to the expected number of records, the ML model is used to determine the actual completion time for processing the expected number of records, and the current number of threads corresponding to the estimated number of threads for the current iteration is stored in the thread and completion time list in association with the actual completion time of the batch job. This involves performing the following repeatedly, Based on the multiple iterations performed for the multiple threads, the threads and a list of completion times are output. The computer execution method according to claim 1, further comprising:

3. Using the aforementioned ML model, determining the actual completion time for processing the expected number of records is: This includes providing the expected number of records to be processed in the current iteration as input to the ML model, wherein the expected number of records is calculated by dividing the total number of records by the current number of threads. The execution time corresponding to the expected number of records is obtained as the output of the ML model, The actual completion time is calculated by adding the thread initialization time to the execution time output by the ML model, The computer execution method according to claim 2, further comprising:

4. The computer execution method further includes selecting a number of maximum efficiency single virtual machines from the cloud virtual machine list such that each of the multiple configurations includes a number of maximum efficiency single virtual machines available in each of the multiple configurations, where each of the multiple configurations corresponds to a number of cores per single virtual machine, and the computer execution method is The computer execution method according to claim 1, further comprising arranging the plurality of maximum efficiency single virtual machines in a list of the plurality of maximum efficiency single virtual machines.

5. The aforementioned combination of threads and completion time lists includes one combination, The aforementioned computer execution method is: The computer execution method includes determining a set of virtual machines comprising a plurality of high-efficiency virtual resources, each of the plurality of high-efficiency virtual resources comprising one of the single virtual machines and virtual machine combinations determined from the plurality of maximum-efficiency single virtual machines, each of the plurality of high-efficiency virtual resources having a number of cores equal to or greater than the total number of cores required to run the batch job within the maximum completion time, and the computer execution method is The aforementioned multiple high-efficiency virtual resources are stored in a high-efficiency virtual resource list in descending order of efficiency, Selecting the optimal virtual machine or the optimal combination of virtual machines from the aforementioned high-efficiency virtual resource list as the most efficient virtual resource, The computer execution method according to claim 4, further comprising:

6. The computer execution method according to claim 5, wherein selecting the optimal virtual machine or the combination of optimal virtual machines as the most efficient virtual resource includes determining the lowest cost of the plurality of highly efficient virtual resources from the highly efficient virtual resource list.

7. The aforementioned combination of threads and completion time lists includes multiple combinations, The aforementioned computer execution method is: For each of the aforementioned combinations, The process further includes iteratively determining a set of virtual machines comprising multiple high-efficiency virtual resources, each of which comprises one of the single virtual machines and virtual machine combinations determined from the list of multiple maximum-efficiency single virtual machines, and each of which comprises a number of cores equal to or greater than the total number of cores required to run the batch job within the maximum completion time. The aforementioned computer execution method is: For each of the aforementioned combinations, The process involves repeatedly saving the aforementioned multiple high-efficiency virtual resources to a high-efficiency virtual resource list in descending order of efficiency, The computer execution method according to claim 4, further comprising selecting the most efficient virtual resource as the optimal virtual machine or the optimal virtual machine combination based on a list of highly efficient virtual resources obtained based on the execution of a number of iterations.

8. This involves generating a training dataset using a certain type of historical batch job and associated training data, By training the input model using the aforementioned training dataset, the ML model for processing a certain type of batch job is generated. It further includes, The computer execution method according to claim 1, wherein the ML model includes a regression model.

9. A computer program product comprising instructions embodied in one or more non-temporary machine-readable storage media and configured to cause one or more data processors to execute the computer execution method described in claim 1.

10. One or more data processors, A non-temporary computer-readable medium that stores instructions causing the one or more data processors to execute the computer execution method described in claim 1 when executed by the one or more data processors, A computer system equipped with the following features.

11. A computer execution method, This includes obtaining a list of threads and completion times for processing a batch job within a maximum completion time, wherein the list of threads and completion times comprises a set of one or more combinations, each of which comprises a thread estimate and an actual completion time for processing the total number of records in the batch job using the thread estimate. The aforementioned computer execution method is: The process further includes selecting the optimal virtual machine or optimal combination of virtual machines from the list of available cloud virtual machines using the thread and completion time list, A computer execution method that enables the execution of the batch job within the maximum completion time, using the aforementioned optimal virtual machine or combination of the aforementioned optimal virtual machines.

12. The selection of the optimal virtual machine or the optimal combination of virtual machines is The process further includes selecting a number of maximum efficiency single virtual machines from the cloud virtual machine list so that each of the multiple configurations includes a maximum efficiency single virtual machine available in each of the multiple configurations, each of which corresponds to the number of cores per single virtual machine, and selecting the optimal virtual machine or the optimal combination of virtual machines. The computer execution method according to claim 11, further comprising arranging the plurality of maximum efficiency single virtual machines in a list of the plurality of maximum efficiency single virtual machines.

13. The aforementioned combination of threads and completion time lists includes one combination, Selecting the aforementioned optimal virtual machine or the aforementioned optimal virtual machine combination is, Determining a set of virtual machines that includes a plurality of highly efficient virtual resources, each containing one of the single virtual machines and virtual machine combinations determined from the list of the plurality of maximum efficiency single virtual machines, and each having a number of cores equal to or greater than the total number of cores required to execute the batch job within the maximum completion time; The aforementioned multiple high-efficiency virtual resources are stored in a high-efficiency virtual resource list in descending order of efficiency, Selecting the optimal virtual machine or the optimal combination of virtual machines from the aforementioned high-efficiency virtual resource list as the most efficient virtual resource, The computer execution method according to claim 12, further comprising:

14. The computer execution method according to claim 13, wherein selecting the optimal virtual machine or the combination of optimal virtual machines as the most efficient virtual resource includes determining one of the plurality of highly efficient virtual resources having the lowest cost from the highly efficient virtual resource list.

15. The aforementioned combination of threads and completion time lists includes multiple combinations, Selecting the aforementioned optimal virtual machine or the aforementioned optimal virtual machine combination is, For each of the aforementioned combinations, The process further includes iteratively determining a set of virtual machines comprising multiple high-efficiency virtual resources, each of which comprises one of the single virtual machines and virtual machine combinations determined from the list of multiple maximum-efficiency single virtual machines, and each of which comprises a number of cores equal to or greater than the total number of cores required to run the batch job within the maximum completion time. Selecting the aforementioned optimal virtual machine or the aforementioned optimal virtual machine combination is, For each of the above multiple combinations The process involves iteratively saving the aforementioned multiple high-efficiency virtual resources to a high-efficiency virtual resource list in descending order of efficiency, The computer execution method according to claim 12, further comprising selecting the most efficient virtual resource as the optimal virtual machine or the optimal virtual machine combination based on a list of highly efficient virtual resources obtained based on the execution of a number of iterations.

16. Obtaining the aforementioned thread and completion time list is: From 2, for each number of threads among the multiple number of threads of the predetermined maximum number of threads, Based on the current number of threads in the current iteration among multiple iterations, the thread initialization time for processing the batch job is determined, The process execution time is determined using the aforementioned maximum completion time and thread initialization time. The aforementioned process execution time is provided as input to the machine learning (ML) model, Based on the execution time of the input process, the number of predicted records to be processed by the current number of threads is obtained as the output of the ML model, The process involves determining whether the predicted number of records is greater than or equal to the expected number of records that can be processed using the current number of threads, If the predicted number of records is greater than or equal to the expected number of records, the ML model is used to determine the actual completion time for processing the expected number of records, and the current number of threads corresponding to the estimated number of threads for the current iteration is stored in the thread and completion time list in association with the actual completion time of the batch job. This involves performing the following repeatedly, Based on the multiple iterations performed for the multiple threads, the threads and a list of completion times are output. The computer execution method according to claim 11, further comprising:

17. Using the aforementioned ML model, determining the actual completion time for processing the expected number of records is: The expected number of records, which is calculated by processing in the current iteration and dividing the total number of records by the current number of threads, is provided as input to the ML model. The execution time corresponding to the expected number of records is obtained as the output of the ML model, The actual completion time is calculated by adding the thread initialization time to the execution time output by the ML model, The computer execution method according to claim 16, including the method described in claim 16.

18. This involves generating a training dataset using a certain type of historical batch job and associated training data, By training the input model using the aforementioned training dataset, the ML model for processing a certain type of batch job is generated. It further includes, The computer execution method according to claim 16, wherein the ML model includes a regression model.

19. A computer program product embodied in one or more non-temporary machine-readable storage media, which includes instructions configured to cause one or more data processors to execute the computer execution method described in claim 11.

20. One or more data processors, A non-temporary computer-readable medium that stores instructions causing the one or more data processors to execute the computer execution method described in claim 11 when executed by the one or more data processors, A computer system equipped with the following features.

Images