Method, system, device and medium for resource allocation based on spatial workload aggregation

Abstract

This invention discloses a resource allocation method, system, device, and medium based on spatial workload aggregation. The method includes: dividing the workload into long-term periodic workloads and other workloads; under the condition of satisfying resource capacity constraints at any given time, and with the objective of minimizing the MCV of the aggregated resource usage sequence in each group of oversold workloads, dividing the long-term periodic workloads into several groups of oversold workloads, where MCV is a coefficient of variation based on the maximum value that characterizes the degree of fluctuation in resource oversold time; for each group of oversold workloads: setting a dynamic allocation coefficient based on its MCV and periodic confidence level, using the product of its historical peak resource usage and the dynamic allocation coefficient as the actual resource allocation amount, and allocating computing resources to the oversold workload group based on the actual resource allocation amount. This invention achieves higher resource utilization, lower overload risk, and a more stable oversold strategy update frequency.

Description

Technical Field

[0001] This invention belongs to the field of cloud computing and resource computing technology, and more specifically, relates to a resource allocation method, system, device and medium based on spatial workload aggregation. Background Technology

[0002] In cloud computing platforms, there is a significant gap between the amount of resources purchased by tenants and the actual amount used, resulting in a large amount of resources remaining idle for extended periods. To improve resource utilization and cloud service provider revenue, cloud platforms typically employ resource over-selling strategies, which allocate logical resource quotas exceeding the actual capacity of physical nodes, thereby reusing idle resources. Existing over-selling solutions mainly fall into two categories: one is a reactive scheduling mechanism triggered by overload events, which alleviates overload by migrating containers or virtual machines, rate limiting, etc., when a node is overloaded; the other is a conservative over-selling strategy based on historical peak values, such as reserving historical peak values ​​or peak values ​​multiplied by a factor (e.g., 1.2 times) for each workload, and only using the remaining resources for over-selling.

[0003] Reactive scheduling mechanisms triggered by overload events rely on node overload detection to trigger reactive mechanisms such as migration, rate limiting, and rescheduling. However, overload detection and handling inevitably involve latency, especially in operations such as container migration, resource reclamation, and load balancing, which often require a time window of several seconds to several minutes. Within this window, nodes may remain in a state of resource contention, leading to increased tenant service response time, decreased throughput, and even triggering cascading timeouts and service avalanches. Conservative overselling strategies based on historical peak usage use historical peak resource usage as the allocation benchmark, such as reserving peak usage for each workload or multiplying the peak usage by a safety factor (e.g., 1.2 times). However, in real production environments, workload peaks are usually caused by brief bursts and are much higher than long-term averages. Therefore, using peak usage as a benchmark leads to excessive long-term resource reservations, significantly compressing the overselling space and resulting in consistently low resource utilization, failing to maximize cloud platform revenue.

[0004] In summary, existing technologies either rely on post-overload remediation, resulting in a poor tenant experience, or are too conservative to improve resource utilization, and lack accurate measurement and optimization mechanisms for workload volatility and peak risk. Therefore, there is an urgent need for a new method that can reduce volatility from the source and support long-term stable overselling. Summary of the Invention

[0005] The main objective of this invention is to provide a resource allocation method, system, device, and medium based on spatial workload aggregation to overcome the shortcomings of the prior art.

[0006] To achieve the above-mentioned objectives, the present invention adopts the following technical solution: The first aspect of this invention provides a resource allocation method based on spatial workload aggregation, comprising: dividing the workload into long-term periodic workloads and other workloads according to the historical resources of the workload; under the condition of satisfying resource capacity constraints at any time, dividing the long-term periodic workload into several groups of oversold workloads with the objective of minimizing the MCV of the aggregated resource usage sequence in each group of oversold workloads after division, wherein MCV is a coefficient of variation index based on the maximum value that characterizes the degree of fluctuation in resource oversold time; for each group of oversold workloads: setting a dynamic allocation coefficient for the oversold workload group according to the MCV and period confidence level of the oversold workload group, using the product of its historical peak resource usage and the dynamic allocation coefficient as the actual resource allocation amount, and allocating computing resources to the oversold workload group according to the actual resource allocation amount.

[0007] Preferably, the workload is divided into long-term periodic workloads and other workloads based on the historical resources of the workload. Specifically, this includes: classifying workloads whose runtime does not exceed a preset threshold as short-term workloads, and classifying workloads whose runtime exceeds the preset threshold as long-term workloads; for each long-term workload: if the period confidence of the long-term workload in the candidate period is greater than the confidence threshold, it is classified as a long-term periodic workload; otherwise, it is classified as a long-term non-periodic workload.

[0008] Preferably, when dividing the long-running workload, the method further includes: denoising the original resource usage sequence of the long-running workload, and calculating an autocorrelation function on the denoised sequence as the periodic confidence of the long-running workload.

[0009] Preferably, the long-term workload is divided into several groups of oversold workloads, specifically including: S21, treating each long-term workload as a group to be divided, forming a set containing multiple groups to be divided; S22, selecting the group to be divided with the largest MCV from the set, and merging the remaining groups to be divided with the group to be divided with the largest MCV one by one; S23, if there is a group to be divided that meets the merging conditions among the remaining groups to be divided, merging the group to be divided with the group to be divided with the largest MCV and placing it into the set, and deleting the group to be divided with the largest MCV from the set; wherein, the merging conditions include: the MCV of the group to be divided and the group to be divided with the largest MCV becomes smaller and meets the capacity constraint; and: the MCV of the group to be divided and the group to be divided with the largest MCV is less than the MCV of other groups to be divided and the group to be divided with the largest MCV; S24, repeating S22-S23 until there are no remaining groups to be divided that meet the merging conditions.

[0010] Preferably, the MCV is: ; in, Describe the MCV of x. This represents the resource usage of workload x at the i-th historical moment. This represents the maximum resource usage of the workload at any given historical time, where N is the total number of historical time points.

[0011] Preferably, the dynamic allocation coefficient is: ; in, The dynamic allocation coefficient is... For tenant request volume, This refers to the historical peak resource usage of the oversold workload group. The periodic confidence level of the oversold workload group. This is the confidence threshold.

[0012] Preferably, the actual resource allocation satisfies: ; in, The actual resource allocation amount, For tenant request volume, This refers to the historical peak resource usage of the oversold workload group.

[0013] A second aspect of the present invention provides a resource allocation system based on spatial workload aggregation, comprising: a first partitioning module, configured to partition the workload into long-term periodic workloads and other workloads based on the historical resources of the workload; a second partitioning module, configured to, under the condition of satisfying resource capacity constraints at any time, partition the long-term periodic workload into several groups of oversold workloads with the objective of minimizing the MCV of the aggregated resource usage sequence in each group of oversold workloads after partitioning, wherein MCV is a coefficient of variation index based on the maximum value characterizing the degree of fluctuation in resource oversold time; and a resource allocation module, configured to, for each group of oversold workloads: set a dynamic allocation coefficient for the oversold workload group based on its MCV and period confidence level, use the product of its historical peak resource usage and the dynamic allocation coefficient as the actual resource allocation amount, and allocate computing resources to the oversold workload group based on the actual resource allocation amount; the resource allocation system based on spatial workload aggregation is used to implement the resource allocation method based on spatial workload aggregation as described above.

[0014] A third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the resource allocation method based on spatial workload aggregation as described above.

[0015] A fourth aspect of the present invention provides a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the resource allocation method based on spatial workload aggregation as described above.

[0016] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention provides a resource allocation method, system, device, and medium based on spatial workload aggregation. It introduces a workload classification and spatial aggregation mechanism based on time volatility, spatially aggregating periodic long-term workloads with peak-shaving characteristics. This effectively reduces the overall time volatility of the aggregated workload without changing the behavior of individual workloads, avoiding the problem of frequent overload and rescheduling caused by drastic fluctuations in individual workloads in traditional overselling methods. Furthermore, it employs an adaptive resource allocation mechanism jointly driven by MCV and periodic confidence, comprehensively considering historical peak resource usage, time volatility, and periodic stability within a unified resource management framework to dynamically determine the actual resource allocation. The proposed over-sales ratio allows the over-sales strategy to automatically adjust according to changes in workload characteristics, avoiding resource waste or stability degradation caused by fixed safety factors or static thresholds. This achieves long-term, stable, high-ratio resource over-sales and improved resource utilization. By using the maximum resource usage value rather than the mean as the volatility assessment benchmark, the invention more accurately characterizes the time volatility features directly related to overload risk in over-sales decisions. Furthermore, it formalizes the workload aggregation problem under multi-dimensional resources into an optimization problem of minimizing the total MCV, thus overcoming the shortcomings of existing methods that rely on mean or variance indicators to guide the design of stable over-sales strategies. Experimental results show that this invention has significant advantages in reducing workload time volatility, improving resource utilization, and enhancing over-sales stability. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating a resource allocation method based on spatial workload aggregation provided in an embodiment of the present invention.

[0019] Figure 2 This is an overall architecture diagram of the resource allocation method based on spatial workload aggregation provided in an embodiment of the present invention.

[0020] Figure 3This diagram illustrates a comparison of the utilization rates of central processing unit (CPU) and memory (MEM) resources between the method of this invention and existing methods.

[0021] Figure 4 This is a schematic diagram comparing the over-selling update ratio of the method of the present invention with that of existing methods.

[0022] Figure 5 This diagram illustrates a comparison of the volatility of the method of this invention with existing methods under high dynamic load scenarios.

[0023] Figure 6 This is a block diagram of a resource allocation system based on spatial workload aggregation, provided in an embodiment of the present invention.

[0024] Figure 7 This is a block diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0025] In view of the shortcomings of the prior art, the inventors of this invention, through long-term research and extensive practice, have proposed the technical solution of this invention. The following will further explain and illustrate this technical solution, its implementation process, and its principles.

[0026] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0027] Furthermore, in the description of this invention, it should be understood that the terms "upper," "lower," "inner," "outer," "horizontal," "vertical," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0028] In the description of this specification, the references to terms such as "an embodiment," "a particular embodiment," or "the embodiment" 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 the invention. 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.

[0029] Figure 1 A flowchart illustrating a resource allocation method based on spatial workload aggregation provided in an embodiment of the present invention. (See also...) Figure 1 , combined Figures 2-5The resource allocation method based on spatial workload aggregation provided in this embodiment of the invention will be described in detail. The method includes operations S1-S3, and its overall framework is as follows: Figure 2 As shown.

[0030] Operation S1 divides the workload into long-term cycle workloads and other workloads based on the historical resources of the workload.

[0031] Operation S2, under the condition of satisfying the resource capacity constraint at any time, aims to minimize the MCV of the aggregated resource usage sequence in each group of oversold workloads after division, and divides the long-term workload into several groups of oversold workloads. MCV is the coefficient of variation based on the maximum value, which characterizes the degree of fluctuation of resource oversold time.

[0032] Operation S3: For each oversold workload group: Set a dynamic allocation coefficient for the oversold workload group based on its MCV and periodic confidence level, use the product of its historical peak resource usage and the dynamic allocation coefficient as the actual resource allocation amount, and allocate computing resources to the oversold workload group based on the actual resource allocation amount.

[0033] The resource allocation method based on spatial workload aggregation provided by this invention is based on the technical route of "periodic identification - spatial aggregation to reduce volatility - adaptive resource allocation based on MCV / ACF". It is mainly committed to achieving long-term stable resource overselling in cloud data centers by reducing the temporal volatility of workloads, reducing the temporal volatility of aggregated workloads, and significantly improving the utilization of CPU and memory resources without reducing the tenant experience, thereby achieving higher resource utilization, lower overload risk and more stable overselling strategy update frequency.

[0034] In a preferred embodiment, operation S1 specifically includes: classifying workloads whose runtime does not exceed a preset threshold as short-term workloads, and classifying workloads whose runtime exceeds a preset threshold as long-term workloads; for each long-term workload: if the period confidence of the long-term workload within the candidate period is greater than the confidence threshold, it is classified as a long-term periodic workload; otherwise, it is classified as a long-term non-periodic workload.

[0035] In a preferred embodiment, when dividing the long-running workload in operation S1, the method further includes: denoising the original resource usage sequence of the long-running workload, and calculating the autocorrelation function on the denoised sequence as the periodic confidence of the long-running workload.

[0036] Specifically, in operation S1, when the monitored workload runtime exceeds a preset threshold (e.g., 72 hours), the workload is classified as a long-term workload; otherwise, it is classified as a short-term workload. For the original resources of long-term workloads, the sequence is denoised (e.g., smoothed using LOESS) to reduce the interference of transient bursts on periodic identification. The autocorrelation function (ACF) is calculated on the denoised sequence as the periodic confidence of the long-term workload. The ACF value of the candidate period (e.g., 24 hours) is compared with the confidence threshold. If the ACF is greater than the confidence threshold, it is determined to be a long-term periodic workload (LPW); otherwise, it is determined to be a long-term non-periodic workload (LNPW). The ACF is used to assess the confidence of the existence of a period, and with a fixed number of sampling points, it can achieve periodic determination with constant time complexity.

[0037] For long-term workloads, this invention introduces the Maximum Value Coefficient Variation (MCV) to characterize the time volatility associated with overselling. MCV uses the maximum resource usage as a benchmark, which differs from the traditional CV which uses the mean as a benchmark, and is more in line with the requirement of using the peak value as a safety baseline for overselling.

[0038] Preferably, the MCV is: ; in, Describe the MCV of x. This represents the resource usage of workload x at the i-th historical moment. This represents the maximum resource usage of the workload at any given historical time, where N is the total number of historical time points.

[0039] In a preferred embodiment, operation S2 employs a max-heap-based iterative merging algorithm to perform aggregation processing, dividing long-term workloads into several groups of oversold workloads, specifically including the following sub-operations S21-S24.

[0040] Sub-operation S21 treats each long-term cycle workload as a group to be divided, forming a set containing multiple groups to be divided.

[0041] Sub-operation S22: Select the group to be partitioned with the largest MCV from the set, and merge the remaining groups to be partitioned with the group to be partitioned with the largest MCV one by one.

[0042] Sub-operation S23: If there is a group among the remaining groups that meet the merging conditions, merge that group with the group with the largest MCV and add it to the set, then delete that group and the group with the largest MCV from the set. The merging conditions include: the merged group with the group with the largest MCV results in a smaller MCV and satisfies the capacity constraint; and: the merged group with the group with the largest MCV results in a smaller MCV than the merged groups of other groups with the largest MCV.

[0043] Sub-operation S24 repeats sub-operations S22-S23 until no more groups meet the merging criteria. The final set of groups to be partitioned represents each oversold workload group, serving as the basic unit for subsequent resource allocation and overselling.

[0044] For short-term and long-term non-periodic workloads, resource allocation is based on tenant request volume to ensure performance and stability. Direct allocation, i.e. For workload groups that can be oversold, resource usage is based on historical peak usage. Determine the actual allocation amount This is used to guide subsequent overselling and scheduling, thereby adaptively balancing stability and utilization. Among these, It needs to monotonically increase with increasing volatility and monotonically decrease with increasing confidence in the cycle.

[0045] To simultaneously reflect both time volatility (MCV) and period confidence (ACF), in a preferred embodiment, the dynamic allocation coefficient is: ; in, For dynamic allocation coefficients, For tenant request volume, This represents the historical peak resource usage of the oversold workload group. For the periodic confidence level of the oversold workload group, This is the confidence threshold.

[0046] In a preferred embodiment, the actual resource allocation satisfies: ; in, This refers to the actual amount of resources allocated. For tenant request volume, This represents the historical peak resource usage of the oversold workload group.

[0047] When allocating resources, the actual amount of resources allocated should be used. It takes oversold workload groups as input, performs deployment and scheduling to achieve node load balancing, and feeds the scheduling results back to the monitoring component to form a closed loop.

[0048] The resource allocation method based on spatial workload aggregation provided by this invention is particularly suitable for cloud computing scenarios, such as for long-term stable resource over-selling systems in cloud data centers, with the operating environment being a cloud data center platform. The data center platform consists of a resource management controller, computing nodes, and a monitoring system. The over-selling system includes a workload classification module, a workload spatial aggregation module, a resource allocation module, and a scheduling module. The working process includes the following steps 1-5.

[0049] Step 1: The user submits a workload deployment request to the cloud data center platform. The monitoring module continuously collects the CPU, memory and other resource usage of each workload during operation, and the resource management controller records the workload's runtime, historical resource usage sequence and current cluster status.

[0050] Specifically, in step 1, the workload submitted by the user is a long-running online service container. The monitoring system collects its CPU and memory usage at fixed time intervals, forming a continuous historical resource usage time series. In this embodiment, a certain workload, running continuously for 10 days, had an average CPU usage of 20% and a peak usage of 80%, exhibiting obvious daily periodic fluctuations.

[0051] Step 2: When the workload runtime exceeds the preset threshold, the resource management controller coordinates with the workload classification module to classify the workload and decides whether to submit it to the workload space aggregation module for further processing based on the classification results.

[0052] Step 3: The workload classification module denoises the historical resource usage sequence of the workload and calculates its autocorrelation function value to determine whether there is a stable periodicity, thereby classifying the workload into short-term workload, long-term non-periodic workload, or long-term periodic workload.

[0053] Specifically, in steps 2 and 3, the workload classification module triggers the classification process after detecting that the workload's runtime exceeds 72 hours. It performs LOESS smoothing and denoising on the historical resource usage sequence and calculates the autocorrelation function (ACF) value over a 24-hour period. When the calculated ACF value exceeds a preset threshold, the workload is classified as a long-term periodic workload; otherwise, it is classified as a long-term non-periodic workload.

[0054] Step 4: The workload spatial aggregation module introduces the maximum value coefficient variation index to quantify the time volatility of long-term workloads, and performs spatial aggregation under the premise of meeting node capacity constraints to build several workload groups to reduce the overall time volatility of the aggregated workload.

[0055] Specifically, in step 4, the workload spatial aggregation module introduces the Maximum Coefficient of Variation (MCV) metric to evaluate the temporal volatility of multiple long-term workloads, and performs spatial aggregation using a max-heap-based iterative merging algorithm. In this embodiment, three workloads with staggered daily cycle characteristics are aggregated into the same workload group. After aggregation, the MCV of the workload group is significantly lower than the MCV of each individual workload, thereby effectively reducing the overall temporal volatility.

[0056] Step 5: The resource allocation module dynamically determines the actual resource allocation based on the historical peak resource usage, time volatility index, and period confidence level of the workload or workload group, and the scheduling module completes the deployment and scheduling of the workload or workload group according to the actual resource allocation.

[0057] Specifically, in step 5, the resource allocation module calculates the dynamic allocation coefficient q based on the historical peak resource usage RMax of the workload group, the maximum value coefficient variation index MCV, and the periodic confidence level ACF, and determines the actual resource allocation amount. When the workload is detected to have low volatility and high periodicity, the allocation coefficient q is set to a smaller value to increase the oversold ratio; when the volatility increases, q is automatically increased to ensure system stability. The scheduling module completes the deployment of workload groups on computing nodes based on the actual resource allocation and continuously receives monitoring feedback during operation to form a closed-loop control.

[0058] The embodiments of this invention demonstrate the proposed method through large-scale simulation experiments and small-scale physical test bench experiments. Figure 2 The performance of the framework was verified and compared with existing cloud resource overselling and scheduling solutions such as Crane, CTVMC, RC-like, and N-sigma.

[0059] See Figure 3The diagram shows a comparison of CPU and memory resource utilization under different schemes. It can be seen that in simulations and test bench experiments driven by real production workloads, compared to existing schemes, this invention significantly improves the overall resource utilization efficiency of the cluster by reducing workload temporal volatility and using aggregated workloads for over-selling and resource allocation. Specifically, in the test bench experiments, this invention achieved a maximum improvement in CPU resource utilization of 64.8% and a maximum improvement in memory resource utilization of 36.7%, effectively alleviating the long-standing problem of resource idleness in cloud data centers.

[0060] See Figure 4 This paper presents a comparison of the oversubscription update ratio under different schemes. By statistically analyzing the oversubscription parameter update ratio triggered by resource overload during the operation of different schemes, the stability of the oversubscription strategy can be measured. Figure 4 The experimental results show that the present invention can significantly reduce the over-sale update frequency. Within the same operating cycle, the over-sale update ratio can be reduced by more than 80% compared with the comparison scheme, thereby reducing the control overhead and performance jitter caused by frequent adjustments and achieving long-term stable resource over-sale.

[0061] See Figure 5 The paper presents a comparison of the volatility of different solutions under high dynamic load scenarios. It can be seen that in the experimental environment with high dynamic factors such as increased workload volatility, changes in resource demand, and changes in node operating status, the resource utilization and over-sales stability of the comparative solutions are significantly degraded. However, the present invention, by reducing the time volatility of the workload itself based on periodic identification and spatial aggregation, maintains a high level of resource utilization improvement and over-sales stability, thus verifying the robustness and engineering feasibility of the present invention in complex production environments.

[0062] Figures 3-5 The experimental results show that the present invention has significant advantages in reducing workload time volatility, improving resource utilization, and enhancing overbooking stability.

[0063] The resource allocation method based on spatial workload aggregation provided in this embodiment of the invention has the following advantages.

[0064] (1) This invention proposes a maximum value coefficient variation index (MCV) for resource overselling scenarios. It uses the maximum value of resource usage instead of the mean as the volatility assessment benchmark, which more accurately describes the time fluctuation characteristics directly related to overload risk in overselling decisions. On this basis, the workload aggregation problem under multi-dimensional resources is formalized into an optimization problem of minimizing the total MCV, thereby overcoming the shortcomings of existing indexes based on mean or variance that are difficult to guide the design of stable overselling strategies.

[0065] (2) This invention introduces a workload classification and spatial aggregation mechanism based on time volatility. By denoising and periodic identification of the historical resource usage sequence of long-running workloads, and further spatially aggregating periodic workloads with peak-shaving characteristics, the overall time volatility of the aggregated workload is effectively reduced without changing the behavior of individual workloads. This avoids the problem of frequent overload and rescheduling caused by the drastic fluctuations of individual workloads in traditional overselling methods.

[0066] (3) The present invention adopts an adaptive resource allocation mechanism driven by MCV and periodic confidence. Under a unified resource management framework, it comprehensively considers historical peak resource usage, time volatility and periodic stability, and dynamically determines the actual resource allocation ratio. This enables the overselling strategy to be automatically adjusted according to changes in workload characteristics, avoiding resource waste or stability decline caused by fixed safety factors or static thresholds, thereby achieving long-term stable high-ratio resource overselling and improved resource utilization.

[0067] Based on the same inventive concept, and corresponding to any of the methods in the above embodiments, the present invention also provides a resource allocation system based on spatial workload aggregation. See also... Figure 6 The resource allocation system 600 based on spatial workload aggregation includes a first partitioning module 610, a second partitioning module 620, and a resource allocation module 630.

[0068] The first partitioning module 610, for example, performs operation S1 to partition the workload into long-term cycle workloads and other workloads based on the historical resources of the workload.

[0069] The second partitioning module 620, for example, performs operation S2, which, under the condition of satisfying the resource capacity constraint at any time, aims to minimize the MCV of the aggregated resource usage sequence in each partitioned group of oversold workloads, and divides the long-term workload into several groups of oversold workloads. MCV is a coefficient of variation index based on the maximum value that characterizes the degree of fluctuation in resource oversold time.

[0070] For example, the resource allocation module 630 performs operation S3. For each group of oversold workloads, the resource allocation module sets a dynamic allocation coefficient for the oversold workload group based on its MCV and periodic confidence level, uses the product of its historical peak resource usage and the dynamic allocation coefficient as the actual resource allocation amount, and allocates computing resources to the oversold workload group based on the actual resource allocation amount.

[0071] The resource allocation system 600 based on spatial workload aggregation is used to implement the above-mentioned resource allocation method based on spatial workload aggregation.

[0072] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, it implements the resource allocation method based on spatial workload aggregation as described in any of the above embodiments.

[0073] Figure 7 This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 710, a memory 720, an input / output interface 730, a communication interface 740, and a bus 750. The processor 710, memory 720, input / output interface 730, and communication interface 740 are interconnected internally via the bus 750.

[0074] The processor 710 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0075] The memory 720 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 720 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 720 and is called and executed by the processor 710.

[0076] The input / output interface 730 is used to connect input / output modules to enable information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.

[0077] The communication interface 740 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0078] Bus 750 includes a pathway for transmitting information between various components of the device, such as processor 710, memory 720, input / output interface 730, and communication interface 740.

[0079] It should be noted that although the above-described device only shows the processor 710, memory 720, input / output interface 730, communication interface 740, and bus 750, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0080] The electronic devices described in the above embodiments are used to implement the resource allocation method based on spatial workload aggregation as described in any of the foregoing embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0081] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, the present invention also provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the resource allocation method based on spatial workload aggregation as described in any of the above embodiments.

[0082] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0083] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the resource allocation method based on spatial workload aggregation as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0084] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention (including the claims) is limited to these examples; within the framework of the invention, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of the invention as described above, which are not provided in the details for the sake of brevity.

[0085] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of the invention, the well-known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided drawings. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of the invention, and this also takes into account the fact that the details of implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of the invention will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuits) have been set forth to describe exemplary embodiments of the invention, it will be apparent to those skilled in the art that the embodiments of the invention may be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0086] Although the invention has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0087] It should be understood that the above embodiments are merely illustrative of the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly. They should not be construed as limiting the scope of protection of the present invention. All equivalent changes or modifications made in accordance with the spirit and essence of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A resource allocation method based on spatial workload aggregation, characterized in that, include: The workload is divided into long-term workloads and other workloads based on its historical resources. Under the condition of satisfying the resource capacity constraint at any time, with the goal of minimizing the MCV of the aggregated resource usage sequence in each group of oversold workloads after division, the long-term periodic workload is divided into several groups of oversold workloads. MCV is a coefficient of variation index based on the maximum value that characterizes the degree of fluctuation in resource oversold time. For each oversold workload group: set a dynamic allocation coefficient for the oversold workload group based on its MCV and periodic confidence, use the product of its historical peak resource usage and the dynamic allocation coefficient as the actual resource allocation amount, and allocate computing resources to the oversold workload group based on the actual resource allocation amount.

2. The resource allocation method based on spatial workload aggregation according to claim 1, characterized in that, Based on historical resources, workloads are categorized into long-term workloads and other workloads, specifically including: Workloads with runtime not exceeding a preset threshold are classified as short-term workloads, and workloads with runtime exceeding the preset threshold are classified as long-term workloads. For each of the aforementioned long-term workloads: if the period confidence of the long-term workload within the candidate period is greater than the confidence threshold, it is classified as a long-term periodic workload; otherwise, it is classified as a long-term non-periodic workload.

3. The resource allocation method based on spatial workload aggregation according to claim 2, characterized in that, When partitioning the long-running workload, the method further includes: The original resource usage sequence of the long-term workload is denoised, and the autocorrelation function is calculated on the denoised sequence as the periodic confidence of the long-term workload.

4. The resource allocation method based on spatial workload aggregation according to claim 1, characterized in that, The long-term workloads are divided into several groups of oversold workloads, specifically including: S21, each of the long-term cycle workloads is treated as a group to be divided, forming a set containing multiple groups to be divided. S22, Select the group to be divided with the largest MCV from the set, and merge the remaining groups to be divided with the group to be divided with the largest MCV one by one; S23, if there is a group to be divided that meets the merging condition among the remaining groups to be divided, merge the group to be divided with the group to be divided with the largest MCV and put it into the set, and delete the group to be divided and the group to be divided with the largest MCV from the set. The merging conditions include: the MCV of the group to be divided and the group to be divided with the largest MCV becomes smaller and the capacity constraint is satisfied after merging; and: the MCV of the group to be divided and the group to be divided with the largest MCV is less than the MCV of other groups to be divided and the group to be divided with the largest MCV after merging. S24, repeat S22-S23 until there are no remaining groups that meet the merging conditions.

5. The resource allocation method based on spatial workload aggregation according to claim 1 or 4, characterized in that, MCV is: ； in, Describe the MCV of x. This represents the resource usage of workload x at the i-th historical moment. This represents the maximum resource usage of the workload at any given historical time, where N is the total number of historical time points.

6. The resource allocation method based on spatial workload aggregation according to claim 1, characterized in that, The dynamic allocation coefficient is: ； in, The dynamic allocation coefficient is... For tenant request volume, This refers to the historical peak resource usage of the oversold workload group. The periodic confidence level of the oversold workload group. This is the confidence threshold.

7. The resource allocation method based on spatial workload aggregation according to claim 1, characterized in that, The actual resource allocation satisfies: ； in, The actual resource allocation amount, For tenant request volume, This refers to the historical peak resource usage of the oversold workload group.

8. A resource allocation system based on spatial workload aggregation, characterized in that, include: The first partitioning module is used to divide the workload into long-term cycle workloads and other workloads based on the historical resources of the workload. The second partitioning module is used to divide the long-term workload into several groups of oversold workloads, with the goal of minimizing the MCV of the aggregated resource usage sequence in each group of oversold workloads after partitioning, under the condition of satisfying the resource capacity constraint at any time. MCV is a coefficient of variation index based on the maximum value that characterizes the degree of fluctuation in resource oversold time. The resource allocation module, for each oversold workload group: the resource allocation module is used to set a dynamic allocation coefficient for the oversold workload group based on its MCV and periodic confidence, use the product of its historical peak resource usage and the dynamic allocation coefficient as the actual resource allocation amount, and allocate computing resources to the oversold workload group based on the actual resource allocation amount. The resource allocation system based on spatial workload aggregation is used to implement the resource allocation method based on spatial workload aggregation as described in any one of claims 1-7.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the resource allocation method based on spatial workload aggregation as described in any one of claims 1-7.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to execute the resource allocation method based on spatial workload aggregation as described in any one of claims 1-7.

Images