CN-122019167-A - Heterogeneous computing power cooperative system and method based on energy efficiency optimization

CN122019167ACN 122019167 ACN122019167 ACN 122019167ACN-122019167-A

Abstract

The invention discloses a heterogeneous computing power coordination system and method based on energy efficiency optimization, which relate to the technical field of heterogeneous computing power coordination and comprise the steps of analyzing abnormal conditions of computing power resources of a regional resource pool in a current period; the method comprises the steps of judging the migration benefit conditions of other resource pools on an abnormal resource pool, analyzing the energy efficiency income conditions of migration tasks between the abnormal resource pool and a target resource pool under different data volumes, acquiring a target task for migrating the abnormal resource pool to the target resource pool according to the task migration data range between the abnormal resource pool and the target resource pool, migrating the target task in the abnormal resource pool to the target resource pool, controlling the target resource pool to cooperatively finish the task in the abnormal resource pool, so that the regional resource pools with different regions with heterogeneous computing power in the cloud platform can effectively cooperatively work, greatly reducing the energy consumed by the regional resource pool for completing the task, and realizing the optimization on the true meaning of energy efficiency.

Inventors

ZHANG PENG
MA ANG
YU HONGJIAN
ZHANG DONGMING

Assignees

光环云数据有限公司

Dates

Publication Date: 20260512
Application Date: 20260130

Claims (10)

1. An energy efficiency optimization-based heterogeneous computing force cooperation method, which is characterized by comprising the following steps of: Step S1, monitoring an area resource pool in a current period to obtain a monitoring record of the area resource pool, acquiring a historical monitoring record of the area resource pool, and analyzing the abnormal condition of the computing power resource of the area resource pool in the current period to obtain an abnormal resource pool; Step S2, acquiring historical migration network records between the abnormal resource pool and other resource pools, and evaluating migration benefit conditions of the other resource pools to the abnormal resource pool by combining the monitoring records of the other resource pools to obtain a target resource pool; Step S3, energy efficiency setting data are obtained from a platform, a target energy efficiency income threshold between an abnormal resource pool and a target resource pool is generated by combining with a monitoring record of the abnormal resource pool, a historical task migration record between the abnormal resource pool and the target resource pool is obtained, energy efficiency data of the abnormal resource pool and the target resource pool are obtained, energy efficiency income conditions brought by migration tasks between the abnormal resource pool and the target resource pool under different data volumes are analyzed, and a task migration data range between the abnormal resource pool and a candidate resource pool is obtained; And S4, acquiring task data of the task to be determined in the abnormal resource pool, acquiring a target task for migrating from the abnormal resource pool to the target resource pool according to a task migration data range between the abnormal resource pool and the target resource pool, migrating the target task in the abnormal resource pool to the target resource pool, and controlling the target resource pool to cooperatively complete the task in the abnormal resource pool.
2. The energy efficiency optimization-based heterogeneous computing power coordination method according to claim 1, wherein the step S1 comprises: step S11, monitoring area resource pools of platforms deployed in different areas in a current period to obtain a monitoring record of the area resource pools in the current period; step S12, acquiring data of each monitoring index of the regional resource pool from the monitoring record; acquiring each historical monitoring record of the regional resource pool, and acquiring the average value and standard deviation of each monitoring index in the regional resource pool from each historical monitoring record; Acquiring a j-th range Q a j =[μ a -j×σ a ,μ a +j×σ a of an a-th monitoring index in a regional resource pool, wherein mu a is the average value of the average values of the a-th monitoring indexes in each historical monitoring record, and sigma a is the standard deviation of the average value of the a-th monitoring index in each historical monitoring record; Acquiring the distance duration between the historical monitoring record and the current period, recording the distance duration as the historical monitoring record, sequencing each historical monitoring record according to the distance duration, acquiring the maximum value t max of the distance duration in each historical monitoring record, and calculating the characteristic time value of the b-th historical monitoring record in each historical monitoring record T b is the distance duration of the b-th history; Step S13, obtaining the sum T sum of characteristic time values of each historical monitoring record, obtaining a plurality of historical monitoring records with average value of a monitoring index in a j-th range Q a j , obtaining the sum T j sum of characteristic time values in the plurality of historical monitoring records, and calculating the characteristic duty ratio of the j-th range Q a j Setting a characteristic duty ratio threshold L △ , and when L j ≥L △ and L j-1 ＜L △ are equal, judging that the j-th range Q a j is the target range of the a-th monitoring index in the regional resource pool in the current period; And acquiring a target range of each monitoring index of the regional resource pool in the current period, and when the maximum value or the minimum value of any one monitoring index in the monitoring record is not in the target range, judging that the regional resource pool has abnormal computing power resources in the current period, and marking the regional resource pool as an abnormal resource pool.
3. The energy efficiency optimization-based heterogeneous computing power coordination method according to claim 1, wherein the step S2 comprises: s21, eliminating abnormal resource pools from each regional resource pool in the platform, and marking the reserved regional resource pool as a target resource pool; Monitoring network states during task migration between the abnormal resource pool and other resource pools to obtain each historical migration network record between the abnormal resource pool and other resource pools; Setting unit time length and acquiring a transmission rate set in a history migration network record; Step S22, obtaining a maximum theoretical network transmission rate B preset between an abnormal resource pool and other resource pools, and calculating a bandwidth utilization rate F d in a d unit time length in the history migration network record; acquiring a mean value F of average bandwidth utilization rate in each history migration network record; Step S23, obtaining a mean mu ́ and a standard deviation sigma ́ of the mean value of the network transmission rates in each unit duration in the transmission rate set recorded by the historical migration network, and calculating a network comprehensive score beta between an abnormal resource pool and other resource pools; When the network comprehensive score beta is larger than a preset threshold value, marking the other resource pools; Step S24, acquiring the marked monitoring records of the other resource pools, and calculating an abnormal risk value gamma g of a g-th monitoring index in the other resource pools; the order m g of obtaining abnormal risk values is used for calculating the risk values of the g-th monitoring index in the other resource pools Wherein M sum is the total number of each other resource pool; And S25, acquiring risk values of all monitoring indexes in other resource pools, judging that the other resource pools have benefits on task migration of the abnormal resource pools when the risk values of all monitoring indexes in the other resource pools are smaller than a preset risk threshold, and marking the marked other resource pools as target resource pools of the abnormal resource pools.
4. The energy efficiency optimization-based heterogeneous computing power coordination method according to claim 3, wherein the step S3 comprises: step S31, acquiring preset energy efficiency setting data from a platform, and acquiring standard values E △ 、W △ and U △ of energy efficiency gain threshold values, network delay and energy efficiency difference values from the energy efficiency setting data; calculating a target energy efficiency gain threshold E ▽ between the abnormal resource pool and the target resource pool; Step S32, acquiring each historical task migration record between an abnormal resource pool and a target resource pool, and establishing an energy consumption linear regression model E total ; collecting the data volume and the total energy consumption of task migration between an abnormal resource pool and a target resource pool in each historical task migration record to obtain a migration data set; According to the migration data set, calculating unit energy consumption k in the energy consumption linear regression model by using a least square method, and calculating fixed energy consumption E fixed according to the unit energy consumption k; Step S33, analyzing energy efficiency income conditions brought by task migration of the abnormal resource pool and the target resource pool under different data volumes in the task migration process, wherein the specific analysis process comprises the following steps: Calculating an energy efficiency benefit value E of task migration from the abnormal resource pool to the target resource pool, and when the energy efficiency benefit value E is larger than a target energy efficiency benefit threshold E ▽ , marking the minimum value of the total data quantity of the task migration from the abnormal resource pool to the target resource pool as a minimum data quantity zeta min of the task migration from the abnormal resource pool to the target resource pool; When the energy efficiency benefit value E is larger than the target energy efficiency benefit threshold E ▽ , the minimum value of the total data amount of the abnormal resource pool to the target resource pool is obtained and is recorded as the minimum data amount zeta min of the abnormal resource pool to the target resource pool for task migration; Step S34, acquiring the task completion rate of a target resource pool to an abnormal resource pool in a historical task migration record; Obtaining a preset maximum total task processing duration T long in the abnormal resource pool, and calculating a completion duration T △ of the target resource pool on the migration task in the abnormal resource pool; And when the T △ <T long is obtained, the maximum value of the total data amount expected to be migrated from the abnormal resource pool to the target resource pool is recorded as the maximum data amount zeta max for the task migration from the abnormal resource pool to the target resource pool, and the task migration data range zeta= [ zeta min, ζ max ] between the abnormal resource pool and the target resource pool is obtained.
5. The energy efficiency optimization-based heterogeneous computing power coordination method according to claim 4, wherein the step S4 comprises: step S41, task data of each undetermined task in an abnormal resource pool is obtained, wherein the task data comprise the total data of the undetermined task; Acquiring each target resource pool of the abnormal resource pool, and acquiring a task migration data range between the abnormal resource pool and each target resource pool; step S42, marking unfinished tasks in an abnormal resource pool as pending tasks, randomly selecting a plurality of pending tasks from the abnormal resource pool according to task migration ranges in the abnormal resource pool and the target resource pool, and taking the tasks as target tasks for task migration from the abnormal resource pool to the target resource pool, wherein the sum of data amounts of the plurality of pending tasks is in the task migration ranges in the abnormal resource pool and the target resource pool; And transferring the target task in the abnormal resource pool to the target resource pool, and controlling the target resource pool to cooperatively finish the task in the abnormal resource pool.
6. An energy efficiency optimization-based heterogeneous computing power coordination system for executing the energy efficiency optimization-based heterogeneous computing power coordination method according to any one of claims 1 to 5, wherein the system comprises a computing power resource anomaly analysis module, a resource pool migration benefit evaluation module, a migration benefit analysis module and a task coordination module; The computing power resource abnormality analysis module is used for analyzing abnormal conditions of computing power resources of the regional resource pool to obtain an abnormal resource pool; the resource pool migration benefit evaluation module is used for acquiring historical migration network records between the abnormal resource pool and other resource pools, evaluating migration benefit conditions of the other resource pools on the abnormal resource pool, and obtaining a target resource pool; The migration profit analysis module is used for generating a target energy efficiency profit threshold between the abnormal resource pool and the target resource pool, acquiring energy efficiency data of the abnormal resource pool and the target resource pool, analyzing energy efficiency profits brought by migration tasks between the abnormal resource pool and the target resource pool under different data volumes, and obtaining a task migration data range between the abnormal resource pool and the candidate resource pool; The task cooperation module is used for acquiring a target task which is migrated from the abnormal resource pool to the target resource pool according to the task migration data range, migrating the target task in the abnormal resource pool into the target resource pool, and controlling the target resource pool to cooperatively finish the task in the abnormal resource pool.
7. The heterogeneous computing power coordination system based on energy efficiency optimization according to claim 6, wherein the computing power resource abnormality analysis module comprises an index range acquisition unit and a computing power resource abnormality analysis unit; the index range acquisition unit is used for acquiring the target range of each monitoring index of the regional resource pool in the current period; The computing power resource abnormality analysis unit is used for analyzing computing power resource abnormality of the regional resource pool in the current period according to the target range of each monitoring index to obtain an abnormal resource pool.
8. The heterogeneous computing power collaborative system based on energy efficiency optimization according to claim 6, wherein the resource pool migration benefit assessment module comprises a resource pool marking unit and a resource pool migration benefit assessment unit; The resource pool marking unit is used for calculating the network comprehensive scores between the abnormal resource pool and other resource pools and marking the other resource pools according to the network comprehensive scores; the resource pool migration benefit evaluation unit is configured to calculate risk values of each monitoring index in the other resource pools, and when the risk values of each monitoring index in the other resource pools are smaller than a preset risk threshold, determine that the other resource pools have benefits for task migration of the abnormal resource pool, and mark the other resource pools as target resource pools of the abnormal resource pool.
9. The energy efficiency optimization-based heterogeneous computing power collaboration system of claim 6, wherein the migration revenue analysis module comprises an energy efficiency revenue threshold generation unit and a migration revenue analysis unit; the energy efficiency profit threshold generating unit is used for acquiring energy efficiency setting data and monitoring records of the abnormal resource pool and generating a target energy efficiency profit threshold between the abnormal resource pool and the target resource pool; the migration profit analysis unit is used for analyzing energy efficiency profit conditions brought by migration tasks between the abnormal resource pool and the target resource pool under different data volumes according to the target energy efficiency profit threshold value to obtain a task migration data range between the abnormal resource pool and the candidate resource pool.
10. The energy efficiency optimization based heterogeneous computing power collaboration system of claim 6, wherein the task collaboration module comprises a task collaboration unit; The task cooperation unit is used for acquiring a target task for transferring the abnormal resource pool to the target resource pool according to a task transfer data range between the abnormal resource pool and the candidate resource pool, transferring the target task in the abnormal resource pool to the target resource pool, and controlling the target resource pool to cooperatively finish the task in the abnormal resource pool.

Description

Heterogeneous computing power cooperative system and method based on energy efficiency optimization Technical Field The invention relates to the technical field of heterogeneous computing power coordination, in particular to an energy efficiency optimization-based heterogeneous computing power coordination system and method. Background In the process of completing the calculation task, no calculation architecture is universal, so that calculation units with different architectures are required to be combined to form strong comprehensive calculation power, heterogeneous calculation power is cooperatively managed based on energy efficiency optimization, the cost of electricity charge can be directly reduced, the return rate of calculation power construction is improved, more services are supported by using fewer energy sources, more powerful heterogeneous calculation power can be deployed under the same architecture space and power consumption limitation, more complex tasks are processed, unnecessary energy consumption is directly reduced, carbon emission is reduced, and the environment is protected. At present, in order to ensure the safety and cost optimization of data, a cloud platform generally builds resource pools in different areas, the computing power structures in the resource pools are different, computing power resources are different, when the resource pools process the data, corresponding computing power resources are generally scheduled to complete tasks according to the actual conditions of the resource pools, but in practice, the data are required to be migrated in different resource pools in the task completion process because of the uncertainty of the resource states of different areas, the phenomenon that the task migration cannot be migrated to the resource pools matched with the resource pools is easily caused, and the problems can cause abnormal heterogeneous computing power cooperation in the resource pools, so that the energy consumption of the resource pools in task processing cannot be reduced, and even the task processing is possibly influenced. Disclosure of Invention The invention aims to provide a heterogeneous computing power cooperative system and method based on energy efficiency optimization, which are used for solving the problems in the prior art. In order to achieve the purpose, the invention provides the technical scheme that the heterogeneous calculation force cooperation method based on energy efficiency optimization comprises the following steps: Step S1, monitoring an area resource pool in a current period to obtain a monitoring record of the area resource pool, acquiring a historical monitoring record of the area resource pool, and analyzing the abnormal condition of the computing power resource of the area resource pool in the current period to obtain an abnormal resource pool; Step S2, acquiring historical migration network records between the abnormal resource pool and other resource pools, and evaluating migration benefit conditions of the other resource pools to the abnormal resource pool by combining the monitoring records of the other resource pools to obtain a target resource pool; Step S3, energy efficiency setting data are obtained from a platform, a target energy efficiency income threshold between an abnormal resource pool and a target resource pool is generated by combining with a monitoring record of the abnormal resource pool, a historical task migration record between the abnormal resource pool and the target resource pool is obtained, energy efficiency data of the abnormal resource pool and the target resource pool are obtained, energy efficiency income conditions brought by migration tasks between the abnormal resource pool and the target resource pool under different data volumes are analyzed, and a task migration data range between the abnormal resource pool and a candidate resource pool is obtained; And S4, acquiring task data of the task to be determined in the abnormal resource pool, acquiring a target task for migrating from the abnormal resource pool to the target resource pool according to a task migration data range between the abnormal resource pool and the target resource pool, migrating the target task in the abnormal resource pool to the target resource pool, and controlling the target resource pool to cooperatively complete the task in the abnormal resource pool. Further, step S1 includes: step S11, monitoring area resource pools of platforms deployed in different areas in a current period to obtain a monitoring record of the area resource pools in the current period; step S12, acquiring data of each monitoring index of the regional resource pool from the monitoring record; acquiring each historical monitoring record of the regional resource pool, and acquiring the average value and standard deviation of each monitoring index in the regional resource pool from each historical monitoring record; Acquiring a j-th range Q aj=[μa-j×σa,μa+j