Processing method

Abstract

Reduce the development costs of learning models. [Solution] The processing method includes: an acquisition step of acquiring job information relating to a machine learning job being performed at one of several nodes; an identification step of identifying other nodes among the several nodes that can complete the remaining part of the job within the learning deadline based on the job information; a comparison step of comparing a first cost of continuing the job at one node with a second cost of performing the remaining part of the job at another node; and a move step of moving the job from one node to another node if the second cost is less than the first cost.

Description

Technical Field

[0001] The present invention relates to a processing method, and more specifically, to the technical field of a processing method for processing jobs related to machine learning.

Background Art

[0002] Services using a trained model (i.e., AI (Artificial Intelligence)) generated by machine learning have been proposed. For example, in Patent Document 1, the types and installation environments of signs or markings around a vehicle, the driving situation of the vehicle, the position of the vehicle, and the line-of-sight direction of the driver of the vehicle are input into the trained model, and a service for providing safe driving support information based on the output of the trained model is described.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the technology described in Patent Document 1, a trained model is used in an in-vehicle device, but a trained model may also be used in a data center having higher processing power than an in-vehicle device. For example, in the service for providing safe driving support information described in Patent Document 1, it is required to process relatively little data from a vehicle in real time. On the other hand, in machine learning for generating a learning model, it is required to process a large amount of data. That is, the performance required for the data center used to provide the above service is different from the performance required for the data center implementing machine learning. By the way, the cost required for developing a learning model often varies depending on the data center used for developing the learning model (in other words, for implementing machine learning). If no countermeasures are taken, there is a technical problem that the development cost of the learning model increases.

[0005] This invention has been made in view of the above-mentioned problems, and aims to provide a processing method that allows for the selection of a data center in a way that suppresses the development costs of a learning model. [Means for solving the problem]

[0006] A processing method according to one aspect of the present invention includes: an acquisition step of acquiring job information relating to a machine learning job being performed at one of a plurality of nodes; an identification step of identifying other nodes among the plurality of nodes that can complete the remaining part of the job within the learning deadline based on the job information; a comparison step of comparing a first cost for continuing the job at the one node with a second cost for performing the remaining part of the job at the other node; and a move step of moving the job from the one node to the other node if the second cost is less than the first cost. [Brief explanation of the drawing]

[0007] [Figure 1] This is a conceptual diagram illustrating the concept of the system according to the embodiment. [Figure 2] This is a block diagram showing the configuration of the information processing device according to the embodiment. [Figure 3] This figure shows an example of computing infrastructure information. [Figure 4] This figure shows an example of job information. [Figure 5] This figure shows an example of an image used to input job information. [Figure 6] This is a conceptual diagram showing an example of job processing according to the embodiment. [Figure 7] This is a flowchart showing an example of job processing according to the embodiment. [Figure 8] This figure shows an example of an image illustrating the results of machine learning. [Modes for carrying out the invention]

[0008] Embodiments relating to the processing method will be described with reference to Figures 1 to 8.

[0009] (system) A system according to an embodiment will be described with reference to Figure 1. In Figure 1, System 1 comprises data centers DC1, DC2, and DC3, and clouds CL1 and CL2, which are connected to each other via a network NW. The number of data centers in System 1 may be two or less, or four or more. The number of clouds in System 1 may be one, or three or more.

[0010] Furthermore, the locations of data centers DC1, DC2, and DC3 are arbitrary. For example, data center DC1 may be in Japan, data center DC2 may be in the United States, and data center DC3 may be in Europe. For example, data center DC1 may be in Aichi Prefecture, data center DC2 may be in Kyushu, and data center DC3 may be in Hokkaido.

[0011] Furthermore, at least one of the data centers DC1, DC2, and DC3 may be a containerized data center. Used batteries from BEVs (Battery Electric Vehicles) may be used as at least part of the power supply for the containerized data center.

[0012] At least one of data centers DC1, DC2, and DC3 may be a data center owned by the company (i.e., an on-premises data center). Data centers DC1, DC2, and DC3 may include data centers provided by other providers (i.e., hosted data centers). At least one of clouds CL1 and CL2 may be a public cloud where the environment built by the cloud provider is shared with other users. Clouds CL1 and CL2 may include a hosted private cloud where a specific user has exclusive use of the cloud environment provided by the cloud provider. Note that a hosted data center and a hosted private cloud may be the same concept.

[0013] Furthermore, data centers DC1, DC2, and DC3, as well as clouds CL1 and CL2, may be referred to as "nodes." The network NW may also be referred to as a "link." Therefore, System 1 can be described as a computing infrastructure comprising multiple nodes configured to communicate via the network NW.

[0014] System 1 includes a database DB. The database DB contains training data used for machine learning. The training data included in the database DB may be training data from a commercially available training dataset. The training data included in the database DB may be training data based on data collected from multiple vehicles (e.g., connected cars).

[0015] In System 1, machine learning using at least a portion of the training data contained in the database DB may be performed in at least a portion of the data centers DC1, DC2, and DC3, as well as in at least a portion of the clouds CL1 and CL2.

[0016] (Configuration of information processing device) System 1 includes an information processing apparatus 100. The information processing apparatus 100 will be described with reference to FIG. 2. In FIG. 2, the information processing apparatus 100 includes an arithmetic unit 110, a storage unit 120, a communication unit 130, an input unit 140, and an output unit 150. The arithmetic unit 110, the storage unit 120, the communication unit 130, the input unit 140, and the output unit 150 may be connected via a data bus 160.

[0017] Note that the information processing apparatus 100 may not include at least one of the input unit 140 and the output unit 150. In this case, at least one of the input unit 140 and the output unit 150 may be connected to the information processing apparatus 100 via an input / output port (not shown) of the information processing apparatus 100 (that is, at least one of the input unit 140 and the output unit 150 may be externally attached to the information processing apparatus 100).

[0018] The arithmetic unit 110 may have one or more processors. The processor may be, for example, at least one of a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit).

[0019] The storage unit 120 may have one or more memories. The memory may be, for example, at least one of a RAM (Random Access Memory), a ROM (Read Only Memory), a hard disk drive, a magneto-optical disk drive, and an SSD (Solid State Drive).

[0020] The communication unit 130 may be capable of communicating with a device external to the information processing apparatus 100. Note that the communication unit 130 may perform wired communication or wireless communication.

[0021] The input device 140 is a device capable of receiving information input to the information processing device 100 from an external source. The input device 140 may include an operating device (e.g., a keyboard, mouse, touch panel, etc.) that can be operated by the user of the information processing device 100. The input device 140 may also include a recording medium reader capable of reading information recorded on a recording medium that can be attached to or detached from the information processing device 100, such as a USB (Universal Serial Bus) memory. When information is input to the information processing device 100 via the communication device 130 (in other words, when the information processing device 100 acquires information via the communication device 130), the communication device 130 may function as an input device.

[0022] The output device 150 is a device capable of outputting information to the outside of the information processing device 100. The output device 150 may output visual information such as characters and images, auditory information such as sounds, or tactile information such as vibrations. The output device 150 may include, for example, at least one of a display, speaker, printer, and vibration motor. The output device 150 may also be capable of outputting information to a recording medium that can be attached to or detached from the information processing device 100, such as a USB memory stick. When the information processing device 100 outputs information via the communication device 130, the communication device 130 may function as an output device.

[0023] The storage device 120 is capable of storing desired data. The storage device 120 may store a computer program that the arithmetic unit 110 will execute. The storage device 120 may also temporarily store data that the arithmetic unit 110 will use temporarily when the arithmetic unit 110 is executing a computer program.

[0024] Furthermore, the computer program may be recorded on a recording medium that is readable by a computer and is not temporary. In this case, the information processing device 100 may read the computer program from the recording medium using a recording medium reading device (not shown). As a result, the computer program may be stored in the storage device 120. Furthermore, at least one of the following may be used as the recording medium: an optical disc, a magnetic medium, a magneto-optical disc, a semiconductor memory, and any other medium capable of storing a program.

[0025] Furthermore, the computer program may be obtained from an external device (not shown) of the information processing device 100 via the communication device 130. In other words, the information processing device 100 may download the computer program via the communication device 130. As a result, the computer program may be stored in the storage device 120.

[0026] The arithmetic unit 110 may, together with the storage device 120 in which the computer program is stored, execute the processing that the information processing device 100 should perform. In other words, the arithmetic unit 110 may, together with the storage device 120 and the computer program stored in the storage device 120, execute the processing that the information processing device 100 should perform. For example, by the arithmetic unit 110 executing the computer program, a logical functional block for executing the processing that the information processing device 100 should perform may be realized within the arithmetic unit 110.

[0027] For example, the arithmetic unit 110 may have, as functional blocks, an acquisition unit 111, a selection unit 112, a determination unit 113, a specific unit 114, a calculation unit 115, and a comparison unit 116. Furthermore, the arithmetic unit 110 may have, as physically implemented processing circuits, an acquisition unit 111, a selection unit 112, a determination unit 113, a specific unit 114, a calculation unit 115, and a comparison unit 116. At least one of the acquisition unit 111, selection unit 112, determination unit 113, specific unit 114, calculation unit 115, and comparison unit 116 may be implemented in a form that combines a logical functional block and a physical processing circuit (i.e., hardware). Details of the acquisition unit 111, selection unit 112, determination unit 113, specific unit 114, calculation unit 115, and comparison unit 116 will be described later.

[0028] The storage device 120 stores computing resource information 121 and job information 122. Computing resource information 121 is information about computing resources available for machine learning. For example, computing resource information 121 may be information about data centers DC1, DC2, and DC3, and clouds CL1 and CL2, respectively. For example, as shown in Figure 3, computing resource information 121 may be information indicating the computing performance, availability, and failure information of each data center. For example, a data center may be represented by information for identifying the data center. For example, computing performance may be represented in FLOPS (Floating-Point Operations Per Second). For example, availability may be represented by the number of available cores.

[0029] Furthermore, the term "data center" in Figure 3 is not limited to data centers DC1, DC2, and DC3, but also includes clouds CL1 and CL2. As mentioned above, data centers DC1, DC2, and DC3, as well as clouds CL1 and CL2, may be referred to as "nodes." For this reason, the computing resource information 121 may also be referred to as node information. In addition to "data center," "computational performance," "availability," and "usage fee," the computing resource information 121 may include other items.

[0030] Job information 122 is information about a job related to machine learning. For example, as shown in Figure 4, job information 122 may be information indicating the learning deadline, dataset, and data volume for each job. For example, the learning deadline may be a date indicating the learning deadline, or it may be the period from the present until the learning deadline. For example, the dataset may be represented by information to identify the dataset used for machine learning. For example, the data volume may be information indicating the amount of data in the dataset. In addition to "job," "learning deadline," "dataset," and "data volume," job information 122 may include other items.

[0031] When a user of the information processing device 100 registers a job, the image 20 shown in Figure 5 may be displayed on the display of the output device 150 as an example. For example, the user may enter the necessary information into at least one of the multiple input fields included in the image 20 via the input device 140. When the user presses the "OK" button included in the image 20 via the input device 140, the information entered by the user is registered in the job information 122.

[0032] For example, information entered in the input field for "Job Name" in Image 20 may be stored in the "Job" field of Job Information 122. For example, information entered in the input field for "Dataset" in Image 20 may be stored in the "Dataset" field of Job Information 122. For example, information entered in the input field for "Learning Deadline" in Image 20 may be stored in the "Learning Deadline" field of Job Information 122. For example, the information processing device 100 may determine the amount of data in the dataset based on the information entered in the input field for "Dataset" in Image 20. The information processing device 100 may store the determined amount of data in the "Amount of Data" field of Job Information 122.

[0033] (Operation of information processing device) The operation of the information processing device 100 will now be explained. First, the process by which the information processing device 100 selects a data center to execute the job included in the job information 122 (i.e., a data center to perform machine learning corresponding to the job) will be explained. Hereafter, "data center" is a concept that includes not only data centers DC1, DC2, and DC3, but also clouds CL1 and CL2.

[0034] The acquisition unit 111 of the arithmetic unit 110 acquires the amount of data and the learning deadline for a job included in the job information 122. The selection unit 112 of the arithmetic unit 110 may calculate the computational performance required to complete a job within the learning deadline based on the acquired amount of data and the acquired learning deadline.

[0035] The selection unit 112 may extract one or more data centers capable of meeting the calculated computational performance requirements (in other words, capable of completing a job within the learning deadline) based on the computational resource information 121. The selection unit 112 selects a data center to perform a job from the extracted one or more data centers in order to minimize the cost required for machine learning. The selection unit 112 may also select one data center to perform machine learning corresponding to a job. The selection unit 112 may also select multiple data centers to perform machine learning corresponding to a job. If the selection unit 112 selects multiple data centers, the determination unit 113 of the computing device 110 determines the training data to be input to each of the multiple data centers based on the dataset related to the job.

[0036] The selection unit 112, via the communication device 130, causes the selected data center to execute a job. For example, the selection unit 112 may register a job in the queue related to the selected data center. At this time, the information processing device 100 may send a dataset related to the job from the database DB to the selected data center based on the job information 122. If the selection unit 112 selects multiple data centers, the information processing device 100 may send learning data related to the job from the database DB to each of the multiple data centers based on the decision result by the decision unit 113. If the selection unit 112 selects a data center to execute a job, the selection unit 112 may register the data center to execute the job in the job information 122.

[0037] Here, data center usage fees vary from data center to data center. On-premises data is relatively inexpensive, while public cloud data is relatively expensive. Hosted data is often more expensive than on-premises data but cheaper than public cloud data.

[0038] For example, if the extracted data servers include both on-premises and public cloud types, the selection unit 112 may select the on-premises type to minimize the cost of machine learning. For example, if the extracted data servers include both hosted and public cloud types, the selection unit 112 may select the hosted type to minimize the cost of machine learning. Furthermore, if the selection unit 112 selects multiple data centers to perform machine learning for a single job, the selection unit 112 may prioritize selecting the on-premises type to minimize the cost of machine learning.

[0039] Next, we will describe the process by which the information processing device 100 moves the remaining portion of a job, which is included in the job information 122, from one data center to another data center.

[0040] As mentioned above, data center usage fees vary from one data center to another. For example, if another data center becomes available with lower usage fees than the data center where a job is being run, the cost of machine learning can be reduced by moving the remaining part of the job from one data center to the other.

[0041] The concept of job transfer processing will be explained with reference to Figure 6. In Figure 6, Data Center DC1 is assumed to be an on-premises data center, and Cloud CL1 is assumed to be a public cloud. The usage fee for Data Center DC1 is assumed to be lower than the usage fee for Cloud CL1.

[0042] At the time the information processing device 100 selects a data center to execute job J2, data center DC1 is assumed to be executing job J1. Therefore, the information processing device 100 will select cloud CL1, which is different from data center DC1, as the data center to execute job J2.

[0043] At time t1 in Figure 6, job J1 is completed. As a result, data center DC1 becomes available to run other jobs. During the period from time t1 to time t2 in Figure 6, the information processing device 100 may determine whether or not to move job J2r from time t2 onwards in cloud CL1 to data center DC1. For example, the information processing device 100 may determine whether data center DC1 can complete job J2r within the learning period based on the amount of data used in job J2r, the time required to send the data to data center DC1, and the learning period for job J2. Note that time tl in Figure 6 corresponds to an example of the learning period for job J2.

[0044] If data center DC1 determines that it can complete job J2r within the learning period, the information processing device 100 may move job J2r to data center DC1. In this case, the information processing device 100 may send information to cloud CL1 indicating that job J2r will be completed at time t2. The information processing device 100 may register job J3, which corresponds to job J2r, in the queue related to data center DC1. As a result, from time t3 in Figure 6, data center DC1 may perform machine learning corresponding to job J3.

[0045] The initial values ​​of the parameters for the machine learning model corresponding to job J3 may be the parameter values ​​of the machine learning model corresponding to job J2 at time t2. Furthermore, at least a portion of the period from time t2 to time t3 in Figure 6 is transmitted to data center DC1. Here, the data used in job J2r may be transmitted from cloud CL1 to data center DC1, or from database DB to data center DC1. Note that CRIU (Checkpoint / Restore in User Space) may be used for the job movement described above.

[0046] For example, the usage fee for Cloud CL1 may be $10 per 10% of Job J2. For example, the usage fee for Data Center DC1 may be $5 per 10% of Job J2. If Cloud CL1 performs all of Job J2, the cost of performing Job J2 is $100.

[0047] For example, at time t2 in Figure 6, 50% of job J2 may be executed. In this case, job J2r corresponds to 50% of job J2. As shown in Figure 6, if job J2r is moved to data center DC1 as job J3, data center DC1 will execute 50% of job J2. In other words, cloud CL1 will execute 50% of job J2, and data center DC1 will execute the remaining 50%. In this case, the cost required to execute job J2 will be $75. In this way, by moving jobs, the cost required to execute jobs (in other words, the cost required for machine learning) can be reduced.

[0048] The job transfer process will be explained with reference to the flowchart in Figure 7. In Figure 7, the acquisition unit 111 of the information processing device 100 may acquire a job from the job information 122 that is being performed in a data center with relatively high usage fees (step S101). For example, in the process of step S101, the acquisition unit 111 may acquire a job with a low progress rate as the above job. This is because it is expected that the cost reduction effect of job transfer will be greater.

[0049] Next, the identification unit 114 of the information processing device 100 may obtain the learning deadline for the first job from the job information 122. Based on the job information 122, the identification unit 114 may estimate the amount of data to be used for the remaining part of the first job. Based on the estimated amount of data, the identification unit 114 may calculate the data transfer time, which is the time required to move the data. Based on the learning deadline for the first job, the estimated amount of data, and the calculated data transfer time, the identification unit 114 may calculate the computational performance required to complete the remaining part of the first job within the learning deadline. Based on the computing resource information 121, the identification unit 114 may identify other data centers that can meet the calculated computational performance (in other words, that can complete the remaining part of the first job within the learning deadline) (step S102). In other words, the identification unit 114 identifies other data centers based on the time required to perform the remaining part of the first job and the data transfer time. The identification unit 114 may identify multiple data centers as other data centers.

[0050] Next, the calculation unit 115 of the information processing device 100 may calculate the first cost and the second cost (step S103). Here, the first cost is the cost of continuing a job in a data center where a job is currently being executed (for example, the cost of executing the entire job in a data center where a job is currently being executed). The second cost is the cost of executing the remaining part of the job in another data center identified by the identification unit 114. The second cost may be the sum of the cost of executing the remaining part of the job in the other data center and the cost of moving the data used for the remaining part of the job. If the identification unit 114 identifies multiple other data centers, the calculation unit 115 may calculate the second cost for each of the multiple other data centers.

[0051] Here, we will explain the costs of moving data used for the remainder of a job. For example, there are often costs involved when retrieving data from a public cloud. Therefore, if the first data center is a public cloud, the cost of moving the data may be the sum of the cost of retrieving the data used for the remainder of the job from the first data center and the communication costs of transmitting that data to the other data center. If there are no costs involved in retrieving data from the first data center, the cost of moving the data may be equal to the communication costs of transmitting the data used for the remainder of the job to the other data center.

[0052] Next, the comparison unit 116 of the information processing device 100 may compare the first cost and the second cost. The comparison unit 116 may determine whether the second cost is less than the first cost (step S104). If, in the process of step S106, it is determined that the second cost is less than the first cost (step S104: Yes), the information processing device 100 may move the job currently running from one data center to another data center (step S105). If, in the process of step S106, it is determined that the second cost is not less than the first cost (step S104: No), the data center may continue running the job (i.e., the job does not need to be moved) (step S106).

[0053] When machine learning corresponding to a job is completed, the information processing device 100 may obtain result information indicating the result of the job from the data center. The information processing device 100 may store this result information in the storage device 120. The user of the information processing device 100 may have the information processing device 100 display the result information via the input device 140. In this case, the information processing device 100 may display the image 30 shown in Figure 8 on a display as an example of an output device 150.

[0054] Furthermore, job migration is not limited to a single instance; it may occur multiple times. For example, a job running on a public cloud may be moved to a hosted private cloud, and then further moved from that hosted private cloud to an on-premises data center. Moreover, job migration is not limited to data centers of different types; it may also occur between data centers of the same type. For example, a job running on a public cloud with relatively high usage fees may be moved to a public cloud with relatively low usage fees.

[0055] For example, when moving data associated with a job, a technique that moves the data necessary for the job's execution (e.g., applications, libraries, dependencies, and at least one of the files) together, such as a Linux Container, may be applied. As mentioned above, there can be costs involved in retrieving data from a data center (e.g., a public cloud). When moving a job running in one data center to another, some of the data related to that job (e.g., training data) may be deleted from the first data center after stopping the job's execution and before moving the data to the other data center. This configuration reduces the amount of data retrieved from the first data center when a job is moved. In other words, it reduces the cost of retrieving data from the first data center. In this case, the data deleted in the first data center may be sent to the other data center, for example, from a database DB. In this case, the data retrieved from the first data center (in other words, the data moved from one data center to another data center) may include metadata about the deleted data.

[0056] (Examples of application) The trained model generated by machine learning using System 1 described above may be applied, for example, to the advanced driver assistance functions (Advanced Drive / Advanced Drive Assistance System) of a vehicle.

[0057] For example, a base model related to advanced driver assistance functions may be generated as a trained model by machine learning using a commercially available training dataset contained in the database DB and System 1. Furthermore, the above base model may be fine-tuned by machine learning using training data based on data collected from multiple vehicles driving in a specific region, contained in the database DB, and System 1. As a result, a training model related to advanced driver assistance functions optimized for a specific region may be generated. LoRA (Low-Rank Adaptation) may be used for fine-tuning.

[0058] (Technical effects) To provide a safer and more comfortable driving environment for vehicles, the use of AI is being considered. For example, a trained model (i.e., AI) related to advanced driver assistance functions can be executed on an in-vehicle system to provide assistance with the operation of peripheral devices such as air conditioners and audio systems, and to support safer driving. In addition to or instead of the in-vehicle system, the trained model related to advanced driver assistance functions can be executed on a server on the network, and more comprehensive services can be provided to the vehicle user via the communication device installed in the vehicle. A server providing such services needs to respond to user requests in real time. However, the data input to such a server is relatively small.

[0059] For example, to develop AI related to advanced driver assistance functions, the server that performs machine learning (corresponding to the servers included in the data center mentioned above) needs to process a large amount of data. However, if the predetermined development schedule is followed, real-time response is not required. In other words, if the predetermined development schedule is followed, the processing time does not need to be short. Thus, the performance requirements for the server that runs the trained model and the server that performs machine learning are different.

[0060] As mentioned above, usage fees often vary depending on the data center. In System 1 according to this embodiment, the information processing device 100 selects a data center such that machine learning is completed within the learning period and the cost of machine learning is minimized. Furthermore, in System 1 according to this embodiment, the information processing device 100 may move a job being performed in one data center to another data center in order to further minimize the cost of machine learning. In other words, in System 1, a data center is selected in such a way that the cost of machine learning is minimized while adhering to a predetermined development schedule. Therefore, according to System 1 according to this embodiment, a data center can be selected in such a way that the development cost of the learning model is suppressed.

[0061] Furthermore, System 1 may include an on-premise data center that reliably meets the company's computing needs, and at least one of a hosted private cloud and / or public cloud that meets the remaining portion of the company's computing needs. The information processing device 100 may select a data center such that machine learning is completed within the learning period and the costs required for machine learning are minimized. With this configuration, the company can meet its computing needs while suppressing the development costs of the learning model.

[0062] (First variation) The computing resource information 121 may further include environmental impact information indicating the environmental burden related to the data center. For example, the environmental impact information may include indicators that show the environmental burden. For example, the environmental impact information may include information indicating the type of energy used by the data center. The type of energy may include, for example, green energy, renewable energy, fossil fuels, etc.

[0063] For example, the calculation unit 115 of the information processing device 100 may calculate the first and second costs based on the environmental load information included in the computing resource information 121. In this case, the first cost may include the first environmental load cost relating to the environmental load of one data center currently performing a job. The second cost may include the second environmental load cost relating to the environmental load of other data centers identified by the identification unit 114. For example, the environmental load cost may be the cost of reducing the environmental load generated by the data center. In this case, the environmental load cost of a data center using fossil fuels may be higher than the environmental load cost of a data center using green energy. With this configuration, it is possible to suppress both the development cost of the learning model and the environmental load.

[0064] (Second variation) The computing resource information 121 may further include environmental impact information indicating the environmental burden related to the data center, and power information regarding the power situation in the area including the data center. The power information may include, for example, the amount of power generated by solar power generation, the amount of power generated by wind power generation, the amount of energy stored in storage batteries, whether or not output curtailment is being implemented, etc.

[0065] The calculation unit 115 of the information processing device 100 may calculate a first score and a second score instead of a first cost and a second cost. The first score is the score when continuing a job in a data center where a job is currently being performed (for example, when performing the entire job in a data center where a job is currently being performed). The second score is the score when performing the remaining part of a job in another data center identified by the identification unit 114. The first score and the second score may be calculated based on a monetary score related to usage fees, an environmental score related to environmental load, and a power score related to power supply status. Here, the monetary score may be smaller the cheaper the usage fees of the data center. The environmental score may be smaller the smaller the environmental load of the data center. The power score may be smaller the larger the power supply surplus of the region including the data center.

[0066] For example, the calculation unit 115 may calculate the first score as “w1 × (first monetary score) + w2 × (first environmental score) + w3 × (first power score)”. Here, the first monetary score, first environmental score, and first power score refer to the monetary score, environmental score, and power score related to one data center. The calculation unit 115 may calculate the second score as “w1 × (second monetary score) + w2 × (second environmental score) + w3 × (second power score)”. Here, the second monetary score, second environmental score, and second power score refer to the monetary score, environmental score, and power score related to other data centers. Also, “w1”, “w2”, and “w3” are weights. Weight w1 is greater than weights w2 and w3. The relative magnitudes of weights w2 and w3 may be determined according to the user's policy.

[0067] The comparison unit 116 of the information processing device 100 may compare the first score and the second score. If the second score is smaller than the first score, the information processing device 100 may move the job currently running from one data center to another data center. If the second score is not smaller than the first score, the data center may continue running the job (i.e., the job does not need to be moved).

[0068] For example, the amount of electricity generated by solar and wind power is susceptible to weather conditions. If the amount of electricity generated by at least one of the solar and wind power sources exceeds the amount used, at least one of them will be temporarily shut down. In other words, there may be cases where solar and wind power cannot be fully utilized. Data centers consume a relatively large amount of electricity. Therefore, if data centers are located in areas with relatively large surplus power supply capacity, it can be expected that opportunities to utilize solar and wind power will increase.

[0069] Therefore, if the decision of whether or not to move a job is made based on the first and second scores, it is possible to reduce the development costs of the learning model while also reducing the environmental impact. In addition, if the job is moved to a data center in an area with relatively large power supply surplus, it is possible to prevent the temporary shutdown of at least one of the solar and wind power generation systems.

[0070] Aspects of the invention derived from the embodiments and modifications described above are described below.

[0071] A processing method according to one aspect of the invention includes: an acquisition step of acquiring job information relating to a machine learning job being performed on one of a plurality of nodes; an identification step of identifying other nodes among the plurality of nodes that can complete the remaining part of the job within the learning deadline based on the job information; a comparison step of comparing a first cost for continuing the job on the one node with a second cost for performing the remaining part of the job on the other nodes; and a move step of moving the job from the one node to the other nodes if the second cost is less than the first cost. In the above embodiment, "data centers DC1, DC2 and DC3, and clouds CL1 and CL2" correspond to examples of "nodes".

[0072] In one example of the processing method, the specific step may identify the other nodes based on the time required to perform the remainder of the job and the time required to move the data used to perform the remainder of the job, calculated based on the job information. In this example, the second cost may be the sum of the cost of performing the remainder of the job and the cost of moving the data.

[0073] In other examples of the processing method, the first cost may include a first environmental burden cost relating to the environmental burden of one node, and the second cost may include a second environmental burden cost relating to the environmental burden of the other node.

[0074] In another example of the processing method, the comparison step may compare, instead of the first cost, a first score for continuing the job at one node with a second score for performing the remainder of the job at the other node; the move step may move the job from one node to the other node if the second score is less than the first score; the first score may be calculated based on the monetary score, environmental impact score and power score for the one node; and the second score may be calculated based on the monetary score, environmental impact score and power score for the other node.

[0075] Another example of the processing method may include an output step that outputs a report about the job when the job is completed, and the report may include information about the costs.

[0076] In other examples of the system, the plurality of nodes may include at least two of the following: on-premises, private cloud, and public cloud.

[0077] A system according to one aspect of the invention is a system for controlling machine learning of a model in a computing infrastructure comprising a plurality of nodes configured to communicate via a network, comprising: acquisition means for acquiring job information relating to a machine learning job being performed on one of the plurality of nodes; identification means for identifying other nodes among the plurality of nodes that can complete the remaining part of the job within the learning deadline based on the job information; and comparison means for comparing a first cost for continuing the job on the one node with a second cost for performing the remaining part of the job on the other node, wherein if the second cost is less than the first cost, the job is moved from the one node to the other node.

[0078] The present invention is not limited to the embodiments described above, and can be modified as appropriate without contradicting the gist or idea of ​​the invention as can be read from the claims and specification as a whole. Processing methods involving such modifications are also included within the technical scope of the present invention. [Explanation of symbols]

[0079] 1...System, 100...Information Processing Device, 111...Acquisition Unit, 112...Selection Unit, 113...Decision Unit, DB...Database, DC1, DC2, DC3...Data Center, CL1, CL2...Cloud, NW...Network

Claims

1. The process involves acquiring job information related to machine learning jobs being performed on one of several nodes, and Based on the job information, the process of identifying other nodes among the plurality of nodes that can complete the remaining part of the job within the learning deadline, A comparison step of comparing the first cost of continuing the job on one node with the second cost of performing the remaining part of the job on the other node, If the second cost is less than the first cost, the process of moving the job from one node to the other node, A processing method that includes this.

2. In the aforementioned specific step, the other nodes are identified based on the time required to perform the remainder of the job and the time required to move the data used to perform the remainder of the job, which are calculated based on the job information. The processing method according to claim 1.

3. The second cost is the sum of the cost of performing the rest of the job and the cost of moving the data. The processing method according to claim 3.

4. The first cost includes the first environmental burden cost relating to the environmental burden on the one node, The aforementioned second cost includes the second environmental burden cost relating to the environmental burden on the other nodes. The processing method according to claim 1.

5. In the comparison step, instead of the first cost, a first score is compared with the case where the job is continued on one node, and instead of the second cost, a second score is compared with the case where the remaining part of the job is performed on the other node. In the aforementioned transfer process, if the second score is smaller than the first score, the job is moved from one node to the other node. The first score is calculated based on the monetary score, environmental impact score, and power score for the node. The second score is calculated based on the monetary score, environmental impact score, and power score of the other nodes. The processing method according to claim 1.

6. The process includes an output step that outputs a report related to the job when the job is completed, The aforementioned report includes information on costs. The processing method according to claim 1.

7. The aforementioned plurality of nodes include at least two of the following: on-premises, private cloud, and public cloud. The processing method according to claim 1.

Images