Cost-aware online computing resource scheduling method and device, terminal and medium

Abstract

The cost-aware online computing resource scheduling method, device, terminal and medium provided by the application, the method comprises: determining a plurality of computing resource services provided by a plurality of computing resource providers based on the current business scenario of the computing resource demand side, generating a scheduling request option corresponding to each computing resource service; determining the computing resource cost required by each scheduling request option; sequentially calling the scheduling request option based on each computing resource cost to execute an online scheduling strategy, wherein the online scheduling strategy comprises a time sequence and a scheduling request option to be selected at each time point on the time sequence. The application generates a scheduling request option corresponding to each computing resource service, sequentially calls the scheduling request option, and executes an online scheduling strategy, thereby realizing optimal on-demand allocation and use of resources, adapting to complex and diversified computing service scenarios, and providing flexible and elastic computing resource scheduling.

Description

Technical Field

[0001] This invention relates to the field of computing resource technology, and in particular to a cost-aware online computing resource scheduling method, apparatus, terminal, and medium. Background Technology

[0002] With the rapid development of emerging businesses such as artificial intelligence, connected vehicles, industrial internet, and metaverse, the demand for computing power is increasing rapidly (computing power includes general-purpose computing power, intelligent computing power, and supercomputing power). Computing resources have become a key resource in future networks, attracting high attention from industry and academia. Among them, with the differentiated business needs led by artificial intelligence, the future network architecture is evolving into a three-tier architecture of cloud-edge-device. Computing resources are shifting from cloud computing, which is traditionally composed of data centers or server clusters, to ubiquitous distribution of cloud-edge-device. Ubiquitous computing power existing in the cloud, edge, and device is connected through the network to achieve efficient sharing. As the demand for computing resources surges, efficient and flexible scheduling of computing resources becomes particularly important.

[0003] Currently, computing power scheduling is mainly responsible for matching the demand side and the supply side of computing power resources. For example, on the demand side, services such as AI large-scale model training bring a large demand for computing power. At the same time, various emerging services on the application side also put forward new requirements for flexible, convenient, and on-demand matching of computing power resources. Through efficient and flexible integrated computing and network scheduling, ensuring user service experience, improving resource utilization, and achieving multi-dimensional optimization of computing power resources in terms of performance, energy consumption, and cost are important issues that future networks need to pay attention to and solve.

[0004] In existing technologies, the value of different computing resources varies greatly depending on the deployment method and deployment node. Current scheduling mechanisms mainly maximize global resource utility through load awareness, ignoring the diversity and differences of computing resources and the diversity of business needs. Therefore, existing computing resource scheduling mechanisms cannot adapt to complex and diversified computing service scenarios.

[0005] Therefore, existing technologies have shortcomings and need to be improved and developed. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a cost-aware online computing resource scheduling method, device, terminal and medium to address the above-mentioned deficiencies of the prior art, and to solve the problem that the computing resource scheduling mechanism of the prior art cannot adapt to complex and diversified computing service scenarios.

[0007] The technical solution adopted by this invention to solve the technical problem is as follows: A cost-aware online computing resource scheduling method, comprising: Based on the current business scenario of computing resource demand, identify multiple computing resource services provided by several computing resource providers, and generate scheduling request options corresponding to each computing resource service. Determine the computing resource cost required for each of the aforementioned scheduling request options; Based on the cost of each computing resource, the scheduling request options are called sequentially to execute the online scheduling strategy. The online scheduling strategy includes a time series and the scheduling request options to be selected at each moment in the time series.

[0008] In one embodiment of this application, determining the computing resource cost required for each of the scheduling request options includes: Obtain the pre-stored unit service capabilities of each computing resource provider, as well as the preset energy consumption factors of each computing resource provider; Based on the unit service capacity and preset energy consumption factor of each computing resource provider, the computing resource cost required for each scheduling request option is obtained.

[0009] In one embodiment of this application, the unit service capability includes at least one of the following: computing power, communication capability for network connection to computing resources, memory capacity and memory bandwidth capability, and external storage capacity and storage bandwidth capability.

[0010] In one embodiment of this application, scheduling request options are sequentially invoked based on the cost of each computing resource to execute an online scheduling strategy, including: Determine the hierarchical relationship between the various scheduling request options; The scheduling request options are invoked sequentially based on the number of times each scheduling request option is invoked. When invoking each scheduling request option, the number of times the scheduling request option is invoked is determined according to the hierarchical relationship and the cost of computing resources. Determine the global contention ratio and, with minimizing the global contention ratio as the optimization objective, execute the online scheduling strategy.

[0011] In one embodiment of this application, determining the hierarchical relationship between the scheduling request options includes: The scheduling request options are sorted based on predetermined factors to obtain the hierarchical relationship of the scheduling request options; the predetermined factors are at least one of computing power resource cost, comprehensive service capability, and business matching degree with the current business scenario.

[0012] In one embodiment of this application, determining the number of times the scheduling request option is invoked based on the hierarchical relationship and computing resource cost includes: Based on the hierarchical relationship, calculate the ratio of the computing resource cost required for the scheduling request option at the adjacent higher level to the computing resource cost required for the scheduling request option at the current level; The ratio of computing power resource costs is used as the strategy factor corresponding to the scheduling request option at the current level, and the number of times the scheduling request option at the current level is called is obtained based on the strategy factor.

[0013] In one embodiment of this application, determining the global contention ratio and executing an online scheduling strategy with minimizing the global contention ratio as the optimization objective includes: Obtain the sum of computing resource costs corresponding to the scheduling request options that have been invoked at the current moment, and determine the optimal offline cost at the current moment; The ratio of the sum of the computing power resource costs to the offline optimal cost is used as the cost bidding ratio of the online scheduling strategy at the current moment; The maximum value among the cost bidding ratios corresponding to each time point in the time series at the current time is taken as the global competition ratio; With the goal of minimizing the global contention ratio, an online scheduling strategy is executed until the service ends or the highest-level scheduling request option is escalated.

[0014] This application also provides a cost-aware online computing resource scheduling device, comprising: The generation module is used to determine multiple computing resource services provided by several computing resource providers based on the current business scenario of the computing resource demand side, and generate scheduling request options corresponding to each computing resource service. The determination module is used to determine the computing resource cost required for each of the scheduling request options; The scheduling module is used to sequentially call scheduling request options based on the cost of each computing resource to execute an online scheduling strategy. The online scheduling strategy includes a time series and the scheduling request options to be selected at each moment in the time series.

[0015] This application also provides a terminal, comprising: a memory, a processor, and a cost-aware online computing resource scheduling program stored in the memory and executable on the processor, wherein the cost-aware online computing resource scheduling program, when executed by the processor, implements the steps of the cost-aware online computing resource scheduling method as described above.

[0016] This application also provides a computer-readable storage medium storing a computer program that can be executed to implement the steps of the cost-aware online computing resource scheduling method described above.

[0017] The present invention provides a cost-aware online computing resource scheduling method, apparatus, terminal, and medium. The method includes: determining multiple computing resource services provided by several computing resource providers based on the current business scenario of the computing resource demand side; generating scheduling request options corresponding to each computing resource service; determining the computing resource cost required for each scheduling request option; and sequentially invoking the scheduling request options based on each computing resource cost to execute an online scheduling strategy. The online scheduling strategy includes a time series and the scheduling request options to be selected at each moment in the time series. This application, by generating scheduling request options corresponding to each computing resource service and sequentially invoking the scheduling request options to execute the online scheduling strategy, achieves optimal on-demand allocation and use of resources, adapts to complex and diverse computing service scenarios, and provides flexible and elastic computing resource scheduling. Attached Figure Description

[0018] Figure 1 This is a flowchart of a preferred embodiment of the cost-aware online computing resource scheduling method of the present invention.

[0019] Figure 2 This is a block diagram of a computing resource scheduling system.

[0020] Figure 3 This is a flowchart of the online computing power scheduling decision-making process of a preferred embodiment of the cost-aware online computing power resource scheduling method in this invention.

[0021] Figure 4 This is a functional principle block diagram of a preferred embodiment of the cost-aware online computing resource scheduling device of the present invention.

[0022] Figure 5 This is a functional principle block diagram of the terminal in this invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0024] This application provides a flexible and efficient computing resource scheduling mechanism, which can provide flexible resource scheduling and allocation schemes for different business needs, thereby realizing the optimal allocation and use of resources on demand, adapting to complex and diversified computing service scenarios, and providing elastic and flexible computing resource scheduling.

[0025] The following description, with reference to the accompanying drawings, outlines a cost-aware online computing resource scheduling method, apparatus, terminal, and medium based on embodiments of this application. Addressing the problem mentioned in the background art that computing resource scheduling mechanisms cannot adapt to complex and diverse computing service scenarios, this application provides a cost-aware online computing resource scheduling method. In this method, based on the current business scenario of the computing resource demand side, multiple computing resource services provided by several computing resource providers are determined, and scheduling request options corresponding to each computing resource service are generated. The computing resource cost required for each scheduling request option is determined. Based on each computing resource cost, the scheduling request options are sequentially invoked to execute an online scheduling strategy. The online scheduling strategy includes a time series and the scheduling request options to be selected at each moment in the time series. By generating scheduling request options corresponding to each computing resource service and sequentially invoking the scheduling request options to execute the online scheduling strategy, this application achieves optimal on-demand allocation and use of resources, adapting to complex and diverse computing service scenarios and providing flexible and elastic computing resource scheduling.

[0026] Please see Figure 1 , Figure 1 This is a flowchart of the cost-aware online computing resource scheduling method in this invention. For example... Figure 1 As shown in the embodiments of the present invention, the cost-aware online computing resource scheduling method includes the following steps: Step S100: Based on the current business scenario of the computing resource demand side, determine multiple computing resource services provided by several computing resource providers, and generate scheduling request options corresponding to each computing resource service.

[0027] The method described in this application can be used in computing resource scheduling engines and platforms, enabling analysis of the current business scenario based on personalized business needs from the computing resource demand side. The scheduling platform of this application possesses diverse and differentiated computing resources, thereby achieving optimal on-demand allocation and use of computing resources. This adapts to complex and diverse computing service scenarios, providing flexible and elastic computing resource scheduling. Specifically addressing the computing resource scheduling problem in AI-driven, diversified computing service scenarios for future networks, it solves issues such as the inability to adapt to personalized business needs and support flexible and elastic computing resource scheduling. Figure 2 As shown, the computing resource scheduling system involved in this application includes: a computing resource demand side, a computing resource provider side, and a computing resource scheduling engine. The computing resource demand side needs to complete personalized business, while the computing resource provider side can provide diversified computing resources. The computing resource provider side is represented as different computing resources according to its own computing hardware, which can be registered with the computing resource scheduling platform or pre-saved as a mapping table.

[0028] Scheduling request options are defined by the computing resource provider. For example, they can be categorized by technical specifications / performance levels (this is the most intuitive way), assuming a computing provider (such as a computing scheduling platform) has a variety of devices ranging from general-purpose CPU servers to high-end GPU clusters. It can define the following scheduling options: Option 1: Basic General Type; Specifications: 4-core CPU, 8GB RAM, 100GB standard cloud disk, 1Gbps network; The corresponding Cap1: average computing power (CPU FLOPS), moderate communication and storage capabilities; Applicable scenarios: Web services, lightweight databases, development and testing environments; Unit cost C1: 0.1 yuan / minute (lowest cost).

[0029] Option 2: High-performance computing type; Specifications: 32-core CPU, 128GB large memory, high memory bandwidth, NVMe local SSD, 10Gbps network; The corresponding Cap2: strong computing power, high memory capacity and bandwidth, and excellent storage I / O; Applicable scenarios: scientific computing, financial analysis, large in-memory databases.

[0030] Unit cost C2: 0.8 yuan / minute.

[0031] Option 3: AI-trained; Specifications: Equipped with 4 NVIDIA A100 GPUs, 96-core CPU, 1TB of memory, ultra-high GPU interconnect bandwidth, and extremely fast local storage; The corresponding Cap3 features extremely high computing power (FLOPS), crucial communication capabilities (between GPUs), and massive memory. Applicable scenarios: Large-scale deep learning model training; Unit cost C3: 20 yuan / minute (very high cost).

[0032] Option 4: Inference-optimized; Specifications: Equipped with 8 dedicated AI inference chips (such as NPU / TPU), low-power CPU, focusing on high-throughput and low-latency inference tasks; The corresponding Cap4: high efficiency in intelligent computing power, but weak general computing power, and the energy consumption factor may be relatively good.

[0033] Applicable scenarios: Online AI model services, real-time video analysis; Unit cost C4: 5 yuan / minute.

[0034] Option 5: Supercomputing type; Specifications: Access to computing nodes in the supercomputing center, equipped with thousand-core CPUs and ultra-high-speed InfiniBand networks; The corresponding Cap5: top-tier supercomputing power, with communication network capabilities as its core; Applicable scenarios: climate simulation, astrophysical calculations, cutting-edge scientific research.

[0035] like Figure 1 As shown, the cost-aware online computing resource scheduling method further includes the following steps: Step S200: Determine the computing resource cost required for each of the scheduling request options.

[0036] In this embodiment of the application, step S200 specifically includes: Step S210: Obtain the pre-stored unit service capacity corresponding to each computing power resource provider, and the preset energy consumption factor of each computing power resource provider; Step S220: Based on the unit service capacity and preset energy consumption factor of each computing resource provider, obtain the computing resource cost required for each scheduling request option.

[0037] Specifically, energy consumption is a major factor in green-driven resource scheduling within computing resources, and this application introduces an energy consumption factor. As an important regulating factor in computing resource scheduling, the energy consumption factor of different computing resources varies due to the differences and diversity of different computing nodes. They are also different; the higher the energy consumption, the higher the cost of computing resources.

[0038] For example, this application is based on the computing power resources of several computing power providers. The computing power resources of different computing power providers vary and differ at different times. These computing power providers can offer m different scheduling request options, such as the first... The scheduling request options have a unit cost of using computing resources of [amount missing]. .

[0039] This application uses the product of the unit service capacity of each computing resource provider and the preset energy consumption factor as the computing resource cost required for each scheduling request option, so as to facilitate online scheduling strategies.

[0040] In one embodiment of this application, the unit service capability includes at least one of the following: computing power, communication capability for network connection to computing resources, memory capacity and memory bandwidth capability, and external storage capacity and storage bandwidth capability.

[0041] Specifically, computing power resources refer to the computing power of hardware devices in future network infrastructure. Computing power includes general-purpose computing power, intelligent computing power, supercomputing computing power, and cutting-edge computing power. Specifically, it mainly includes CPUs, GPUs, NPUs, TPUs, FPGAs, and programmable network devices.

[0042] For computing resource providers, the common unit of measurement for computing resources is FLOPS, which is the number of floating-point operations performed per second. Furthermore, the varying execution efficiency and capabilities of different services depend not only on the floating-point computing power of the computing nodes, but also on factors such as input / output performance, cache speed, and cache size. Therefore, in computing resource scheduling, the unit service capability of computing resource scheduling is expressed as... ,in, This represents the computing power of computing resources. This indicates the network's communication capabilities when connected to computing resources. This indicates the memory capacity and memory bandwidth capability of the computing resource node. This indicates the external storage capacity, storage bandwidth, and other capabilities of the computing resource node.

[0043] This application uses a comprehensive indicator value that fully reflects the service support capability of the unit through the use of its service capabilities.

[0044] like Figure 1 As shown, the cost-aware online computing resource scheduling method further includes the following steps: Step S300: Based on the cost of each computing resource, the scheduling request options are called sequentially to execute the online scheduling strategy. The online scheduling strategy includes a time series and the scheduling request options to be selected at each time point in the time series.

[0045] Specifically, due to the varying needs of different business scenarios, most computing resource demanders cannot accurately predict their computing resource scheduling requirements µ (the number of times a unit of computing resource service is obtained). This application aims to optimize computing resource scheduling costs and obtain µ times of unit computing resource service at the lowest cost, proposing a cost-aware online scheduling strategy.

[0046] In this embodiment of the application, step S300 specifically includes: Step S310: Determine the hierarchical relationship between the scheduling request options; Step S320: Invoke the scheduling request options sequentially based on the number of times each scheduling request option is invoked. When invoking each scheduling request option, determine the number of times the scheduling request option is invoked according to the hierarchical relationship and computing resource cost. Step S330: Determine the global contention ratio and execute the online scheduling strategy with the goal of minimizing the global contention ratio.

[0047] Specifically, this application models computing power scheduling from a cost perspective. The computing power resource scheduling system offers several resource scheduling options, including leasing or purchasing, with a total of m different options. The computing power resource scheduling system, for the business S on the computing power resource demand side, converts the computing power resource services provided by the computing power demand provider into scheduling request options. The number of times a unit of computing power resource service can be obtained for each scheduling request option is: (If one unit of computing power is equivalent to one hour, that is the basic unit), the cost of computing power resources paid is... For online problems, since the number of computing resource service sessions (µ) required by the business cannot be known in advance, after each acquisition of computing resource service, it is possible to continue acquiring it or not to continue acquiring it for other reasons, but all the costs incurred so far cannot be refunded. Therefore, this application must select a computing resource service scheduling plan online.

[0048] This application aims to minimize the global competition ratio. Based on the number of times each scheduling request option is called, the scheduling request options are called sequentially, realizing a pragmatic, gradual, and low-risk online scheduling strategy. That is, without relying on complex prediction models, it finds the long-term optimal computing resource option for each differentiated business in the real cost environment of business operation through sequential trial and error, and finally locks it in, thereby minimizing the total cost over the entire cycle.

[0049] In one embodiment of this application, step S310 specifically involves: sorting each scheduling request option based on predetermined factors to obtain the hierarchical relationship of each scheduling request option; the predetermined factors are at least one of computing power resource cost, comprehensive service capability, and business matching degree with the current business scenario.

[0050] For example, the scheduling request options can be sorted in ascending order of computing resource cost to obtain a hierarchical relationship; or, the scheduling request options can be sorted in ascending order of comprehensive service capability to obtain a hierarchical relationship; or, the scheduling request options can be sorted in ascending order of business matching degree with the current business scenario.

[0051] Specifically, in this application, at each scheduling moment (e.g., every minute), a scheduling request option must be selected to obtain the corresponding service capabilities to support business operations. The hierarchical relationship between these scheduling request options is determined, thereby ensuring that the behavior of the business upgrade options is meaningful. The hierarchical relationship between the number of service requests per unit of computing power that can be obtained by the scheduling request options is represented as follows: ,like This indicates the purchase of computing power resources, thus becoming a computing power owner. This application order... The value is set to a very large value, ensuring that the optimal solution will not select this scheduling request option. Under this condition, this application sets... That is, the scheduling request options are sorted in ascending order of computing power resource cost.

[0052] This application sorts each scheduling request option according to predetermined rules, so as to schedule them sequentially according to the hierarchical relationship obtained by sorting, thereby realizing a reasonable online scheduling strategy.

[0053] In this embodiment of the application, the step S320 of "determining the number of times the scheduling request option is called based on the hierarchical relationship and computing resource cost" specifically means: Step S321: According to the hierarchical relationship, calculate the ratio of the computing resource cost required by the scheduling request option of the adjacent higher level to the computing resource cost required by the scheduling request option of the current level; Step S322: Use the ratio of computing power resource costs as the strategy factor corresponding to the scheduling request option at the current level, and obtain the number of times the scheduling request option at the current level is called based on the strategy factor.

[0054] Specifically, to improve the overall utilization of computing resources during scheduling and reduce resource fragmentation, scheduling options will be... The cost factor is expressed as In other words, the larger the computing resource service required for a given scheduling, the higher the utilization rate of the scheduling resources and the lower the cost, thus demonstrating cost rationality. Furthermore, to improve the feasibility and rationality of cost-aware online scheduling, scheduling strategies... Introducing strategy factors It is a positive integer and has cost regularity, that is, the cost of the next scheduling option is an integer multiple of the cost of the previous option.

[0055] In this way, each online strategy provides a time series and the scheduling options to be selected at each time point in the series, specifically, such as Figure 3 As shown, the online strategy of this application For: Based on the computing resources provided by the computing resource provider, and for the differentiated business S of the computing resource demander, the computing resource demander's choice of scheduling options. From 1 to Scheduled sequentially Next, select the scheduling option. Each scheduling moment is a necessary scheduling moment, meaning that the computing resource demander must choose a scheduling strategy to acquire computing resources to meet the business needs of the user. (For option i, the system has already accumulated calls before entering option i.) When the cumulative usage time or number of times reaches Only when the request switches from option i to option i+1. In other words, after calling a resource request option, if the business requirement is met, there is no necessary point in time, and the business ends; if the business requirement is still not met, it is escalated to the next higher-level scheduling request option; if the business requirement is still not met, the last option m is finally scheduled. The logic of this application is: first try a small package, and if it is insufficient, continue with a small package; if after multiple attempts it is found that the requirement persists, it is escalated to a medium package; if it is still insufficient, it is finally escalated to a buyout.

[0056] This application first calculates the ratio of the computing resource cost required for scheduling request options based on adjacent higher levels to the computing resource cost required for scheduling request options at the current level. The ratio of computing resource costs is used as a strategy factor, and the strategy factor is used as the number of times the scheduling request option at the current level is called. This improves the overall utilization rate of computing resources during scheduling and reduces resource fragmentation.

[0057] In one embodiment of this application, step S330 specifically includes: Step S331: Obtain the sum of computing resource costs corresponding to the scheduling request options that have been invoked at the current time, and determine the optimal offline cost at the current time; Step S332: The ratio of the sum of computing power resource costs to the offline optimal cost is used as the cost bidding ratio of the online scheduling strategy at the current moment; Step S333: Take the maximum value of the cost bidding ratios corresponding to each time point in the time series at the current time as the global competition ratio; Step S334: With minimizing the global contention ratio as the optimization objective, execute the online scheduling strategy until the service ends or the highest level of scheduling request option is upgraded.

[0058] Specifically, this application defines The minimum cost required for scheduling at time t (i.e. the optimal solution for obtaining computing power services for the offline problem at time t). It is an unknown. For any online scheduling strategy , Indicates online scheduling strategy Next, proceed with the acquisition. All costs incurred during a single computing power service are expressed as follows: .

[0059] in, For online scheduling strategies exist Cost-per-moment ratio For online scheduling strategies The global competition ratio, where for any , can be defined as An online strategy for increasing the global contention ratio. The global contention ratio is the maximum value achieved over the domain of time t, and this application uses the global contention ratio as the optimization objective. The range of values ​​for t is all non-negative integers, which is an infinite set. To obtain the global contention ratio, its range needs to be narrowed down to several possible values. Furthermore, regarding the online scheduling strategy... , For strategy The moment when the scheduling request option is selected in the i-th round, i.e., the global contention ratio, will only be reached at the contention ratio value at the moment the scheduling request option is purchased. In all online strategies, assuming that each scheduling request option is a necessary time point, i.e., at which point it is impossible to continue acquiring computing resources without scheduling any options (i.e., the critical point where computing resources are exhausted and scheduling must be performed again), if an online scheduling strategy... satisfy Then it becomes a strategy. This application employs an optimal online scheduling strategy. Through mathematical derivation, it determines the number of repetitions for each stage to ensure that, at any point when usage ceases, the extra cost incurred by the user (compared to the cost of the optimal solution whose duration was known from the outset) is kept within an acceptable proportion (competition ratio).

[0060] This application can provide flexible resource scheduling and allocation schemes for different business needs, thereby realizing the optimal allocation and use of resources on demand, adapting to complex and diversified computing power service scenarios, and providing elastic and flexible computing power resource scheduling.

[0061] In one specific embodiment, suppose a computing resource provider offers multiple scheduling request options, and a user needs to process a computing task with an unknown total duration (in "computation hours," where H is the number of times a unit of computing resource service is provided, and 1 time is equivalent to 1 hour). The user's goal is to complete the computing task at the lowest cost, but since the total duration of the task is unknown, it is necessary to decide online which computing resource scheduling method to use.

[0062] Option descriptions: Scheduling request option 1 (on-demand computing resource instance): cost C1 = 1 yuan / hour, usage permission count H1 = 1 (1 time is equivalent to 1 hour, equivalent to 1 yuan / 1 hour each time).

[0063] Scheduling request option 2 (Reserved computing power resource instance - small package): Cost C2 = 10 yuan, usage permission times H1 = 20 (equivalent to 0.5 yuan / 1 hour).

[0064] Scheduling request option 3 (Reserved computing power resource instance - large package): Cost C2 = 20 yuan, number of times of use H2 = 45 (equivalent to 0.45 yuan / 1 hour).

[0065] Scheduling request option 4 (permanent purchase): Price C4=∞ (extremely high purchase cost, never to be used), therefore scheduling request option 3 is considered the maximum option.

[0066] Cost-aware online scheduling strategy: Purchase in the order of scheduling request options 1 and 2. Next, purchase scheduling request option 3. Each purchase occurs at a "necessary moment," which is the moment when the computing task needs to continue and new computing resources must be acquired before the computing power is exhausted.

[0067] Specific online scheduling sequence for computing resource services: First, schedule option 1: when cumulative usage reaches the threshold. Then scheduling option 2: Z2=20 / 10=2 times; finally scheduling option 3 (reserved computing power resource instance - large package).

[0068] Furthermore, such as Figure 4 As shown, based on the above-described cost-aware online computing resource scheduling method, the present invention also provides a cost-aware online computing resource scheduling device, comprising: The generation module 100 is used to determine multiple computing resource services provided by several computing resource providers based on the current business scenario of the computing resource demand side, and generate scheduling request options corresponding to each computing resource service. The determining module 200 is used to determine the computing resource cost required for each of the scheduling request options; The scheduling module 300 is used to sequentially call scheduling request options based on the cost of each computing resource to execute an online scheduling strategy. The online scheduling strategy includes a time series and the scheduling request options to be selected at each time point in the time series.

[0069] It should be noted that the foregoing explanation of the embodiment of the cost-aware online computing resource scheduling method also applies to the cost-aware online computing resource scheduling device of this embodiment, and will not be repeated here.

[0070] This invention discloses a cost-aware online computing resource scheduling device. Based on the current business scenario of the computing resource demand side, it identifies multiple computing resource services provided by several computing resource providers and generates scheduling request options corresponding to each computing resource service. It then determines the computing resource cost required for each scheduling request option and sequentially calls the scheduling request options based on each computing resource cost to execute an online scheduling strategy. The online scheduling strategy includes a time series and the scheduling request options to be selected at each moment in the time series. By generating scheduling request options corresponding to each computing resource service and sequentially calling the scheduling request options to execute the online scheduling strategy, this application achieves optimal on-demand allocation and use of resources, adapting to complex and diverse computing service scenarios and providing flexible and elastic computing resource scheduling.

[0071] Figure 5 A schematic diagram of the structure of a terminal provided in an embodiment of this application. The terminal may include: The memory 501, the processor 502, and the computer program stored on the memory 501 and capable of running on the processor 502.

[0072] When the processor 502 executes the program, it implements the cost-aware online computing resource scheduling method provided in the above embodiments.

[0073] Furthermore, the terminal also includes: Communication interface 503 is used for communication between memory 501 and processor 502.

[0074] The memory 501 is used to store computer programs that can run on the processor 502.

[0075] The memory 501 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0076] If the memory 501, processor 502, and communication interface 503 are implemented independently, they can be interconnected via a bus to communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, only one line is used in the diagram, but this does not imply that there is only one bus or one type of bus.

[0077] Optionally, in a specific implementation, if the memory 501, processor 502, and communication interface 503 are integrated on a single chip, then the memory 501, processor 502, and communication interface 503 can communicate with each other through an internal interface.

[0078] Processor 502 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0079] This embodiment also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the cost-aware online computing resource scheduling method described above.

[0080] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0081] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0082] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0083] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can read and execute instructions from or in conjunction with such an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). In addition, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically by optically scanning paper or other media, then editing, interpreting or otherwise processing them as necessary, and then storing them in computer memory.

[0084] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0085] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by a program instructing related hardware, and the program can be stored in a computer-readable storage medium. When executed, the program includes one or a combination of the steps of the method embodiments.

[0086] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0087] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A cost-aware online computing resource scheduling method, characterized in that, include: Based on the current business scenario of computing resource demand, identify multiple computing resource services provided by several computing resource providers, and generate scheduling request options corresponding to each computing resource service. Determine the computing resource cost required for each of the aforementioned scheduling request options; Based on the cost of each computing resource, the scheduling request options are called sequentially to execute the online scheduling strategy. The online scheduling strategy includes a time series and the scheduling request options to be selected at each moment in the time series.

2. The cost-aware online computing resource scheduling method according to claim 1, characterized in that, Determining the computational resource cost required for each of the scheduling request options includes: Obtain the pre-stored unit service capabilities of each computing resource provider, as well as the preset energy consumption factors of each computing resource provider; Based on the unit service capacity and preset energy consumption factor of each computing resource provider, the computing resource cost required for each scheduling request option is obtained.

3. The cost-aware online computing resource scheduling method according to claim 2, characterized in that, The unit service capabilities include at least one of the following: computing power, communication capabilities for connecting to computing resources via network, memory capacity and memory bandwidth capabilities, and external storage capacity and storage bandwidth capabilities.

4. The cost-aware online computing resource scheduling method according to claim 1, characterized in that, Based on the cost of each computing resource, the scheduling request options are invoked sequentially to execute the online scheduling strategy, including: Determine the hierarchical relationship between the various scheduling request options; The scheduling request options are invoked sequentially based on the number of times each scheduling request option is invoked. When invoking each scheduling request option, the number of times the scheduling request option is invoked is determined according to the hierarchical relationship and the cost of computing resources. Determine the global contention ratio and, with minimizing the global contention ratio as the optimization objective, execute the online scheduling strategy.

5. The cost-aware online computing resource scheduling method according to claim 4, characterized in that, Determine the hierarchical relationship between the various scheduling request options, including: The scheduling request options are sorted based on predetermined factors to obtain the hierarchical relationship of the scheduling request options; the predetermined factors are at least one of computing power resource cost, comprehensive service capability, and business matching degree with the current business scenario.

6. The cost-aware online computing resource scheduling method according to claim 4, characterized in that, The number of times the scheduling request option is invoked is determined based on the hierarchical relationship and computing resource cost, including: Based on the hierarchical relationship, calculate the ratio of the computing resource cost required for the scheduling request option at the adjacent higher level to the computing resource cost required for the scheduling request option at the current level; The ratio of computing power resource costs is used as the strategy factor corresponding to the scheduling request option at the current level, and the number of times the scheduling request option at the current level is called is obtained based on the strategy factor.

7. The cost-aware online computing resource scheduling method according to claim 4, characterized in that, Determine the global contention ratio, and with minimizing the global contention ratio as the optimization objective, execute an online scheduling strategy, including: Obtain the sum of computing resource costs corresponding to the scheduling request options that have been invoked at the current moment, and determine the optimal offline cost at the current moment; The ratio of the sum of the computing power resource costs to the offline optimal cost is used as the cost bidding ratio of the online scheduling strategy at the current moment; The maximum value among the cost bidding ratios corresponding to each time point in the time series at the current time is taken as the global competition ratio; With the goal of minimizing the global contention ratio, an online scheduling strategy is executed until the service ends or the highest-level scheduling request option is escalated.

8. A cost-aware online computing resource scheduling device, characterized in that, include: The generation module is used to determine multiple computing resource services provided by several computing resource providers based on the current business scenario of the computing resource demand side, and generate scheduling request options corresponding to each computing resource service. The determination module is used to determine the computing resource cost required for each of the scheduling request options; The scheduling module is used to sequentially call scheduling request options based on the cost of each computing resource to execute an online scheduling strategy. The online scheduling strategy includes a time series and the scheduling request options to be selected at each moment in the time series.

9. A terminal, characterized in that, include: The system includes a memory, a processor, and a cost-aware online computing resource scheduling program stored in the memory and executable on the processor. When executed by the processor, the cost-aware online computing resource scheduling program implements the steps of the cost-aware online computing resource scheduling method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that can be executed to implement the steps of the cost-aware online computing resource scheduling method as described in any one of claims 1 to 7.

Images