CN115333962A - Cloud rendering container resource scheduling method based on fairness and efficiency balance - Google Patents
Cloud rendering container resource scheduling method based on fairness and efficiency balance Download PDFInfo
- Publication number
- CN115333962A CN115333962A CN202210872619.6A CN202210872619A CN115333962A CN 115333962 A CN115333962 A CN 115333962A CN 202210872619 A CN202210872619 A CN 202210872619A CN 115333962 A CN115333962 A CN 115333962A
- Authority
- CN
- China
- Prior art keywords
- container
- task
- cloud
- tasks
- acquired
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/50—Network service management, e.g. ensuring proper service fulfilment according to agreements
- H04L41/5003—Managing SLA; Interaction between SLA and QoS
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/104—Peer-to-peer [P2P] networks
- H04L67/1074—Peer-to-peer [P2P] networks for supporting data block transmission mechanisms
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Automatic Analysis And Handling Materials Therefor (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention relates to the technical field of cloud computing, in particular to a cloud rendering container resource scheduling method based on fairness and efficiency balance.
Description
Technical Field
The invention relates to the technical field of cloud computing, in particular to a cloud rendering container resource scheduling method based on fairness and efficiency balance.
Background
In a cloud computing environment, service providers and users agree on QoS (Quality of Service) standards in SLA (Service level Agreement). However, both parties have different goals, with the service provider focusing on system efficiency and the user focusing on the expected completion time. Therefore, a cloud service provider needs to formulate a fair and efficient resource allocation method, so that on one hand, the differentiated requirements of user task QoS defined by multiple levels of SLAs are met, the fair resource allocation aiming at the differentiated requirements of users is realized, and the satisfaction degree of the users is improved; on the other hand, the system efficiency is improved as much as possible, the optimal balance between the task completion time and the system efficiency is realized, and more economic benefits are obtained.
However, the diversity of user goals, the scalability of the cloud architecture, and the parallel-ready nature of the rendering task make it difficult to trade off fairness and efficiency. The research targets of the existing resource scheduling method are usually focused on the overall performance of the application, such as the average completion time, the blocking rate and the like of tasks, and the problem that different users have different requirements on the QoS of the tasks is rarely considered, so that the problem of fair resource allocation aiming at the differentiated QoS requirements in the cloud environment cannot be solved. Currently, there is no consensus that how the concept of fair allocation of cloud resources should be defined, and traditional fair resource allocation research mainly focuses on the requirements of different types of resources, for example, how to fairly allocate heterogeneous resources to users. However, the fair allocation mechanism of heterogeneous resources is difficult to perform performance optimization by fully utilizing the scalability of cloud resources and the parallelism of rendering tasks, which leads to problems of low rendering efficiency, resource waste, and the like.
According to the parallel-easy characteristic of the rendering process, the rendering task can be evenly divided into subtasks with the same workload, and the execution time of the rendering task is determined by the last completed subtask. According to the law of wooden barrels, the wooden board with the shortest wooden barrel determines the water storage capacity of the wooden barrel, and all the wooden boards need to be kept at a higher height if the water storage capacity of the water barrel needs to be improved, so that in order to reduce the completion time of a rendering task, a cloud service provider preferably uses a homogeneous server to provide services for subtasks with the same workload. However, the complexity of the conventional cloud environment and the heterogeneity of the cloud resources make it difficult for the service provider to fully utilize the parallel-easy feature of the rendering task, which greatly hinders the efficiency improvement of the rendering task. In recent years, the cloud-native concept is gradually mature, the container technology of the cloud-native concept effectively solves the problem of deployment consistency of heterogeneous environments, and heterogeneous resource standardization is achieved. Therefore, the rendering application is deployed based on the cloud native technology, the greedy strategy-based hyper-heuristic algorithm is fused with the multi-objective constraint function, the scalability and the standard of the cloud container and the parallel-easy characteristic of the rendering task are fully utilized, the cloud rendering container resource scheduling method and system with the balance between fairness and efficiency are formulated, the service quality and the rendering efficiency are improved, and the fair and efficient rendering service is realized.
Disclosure of Invention
The invention provides a cloud rendering container resource scheduling method based on fairness and efficiency balance, aiming at making up the defect that the efficiency of a cloud rendering service and the user satisfaction degree are difficult to balance in the prior art, the deployment and the management of rendering application are carried out based on a cloud native technology, a greedy strategy-based hyper-heuristic algorithm is fused with a multi-target constraint function, a fair and efficient cloud rendering task and resource allocation method and system are respectively formulated under the conditions of resource competition and non-competition, and the high efficiency of the cloud rendering service is realized while the user satisfaction degree is improved.
A cloud rendering container resource scheduling method based on fairness and efficiency balance comprises the following steps:
s1, deploying rendering services based on a cloud native technology, collecting rendering task feature types, user requirements and submission time data through a cloud rendering resource management system, analyzing task workloads and weighted values by using the collected data, and summarizing all task requirements;
s2, collecting available cloud containers of the rendering service through a cloud rendering resource management system, analyzing the computing capacity of the available containers, and summarizing all the available cloud containers;
s3, judging whether the existing resources can meet the task computing requirements of all users or not based on the acquired total task requirements and the computing capacity of the existing available cloud container;
s4, if the judgment result of the S3 is positive, determining an optimal resource allocation scheme by adopting a fair efficiency balancing method under resource non-competition based on the acquired available cloud container data and the acquired task characteristic data including task workload and weight;
s5, if the judgment result of the S3 is negative, determining an optimal resource allocation scheme by adopting a fair efficiency balancing method under resource competition based on the acquired available cloud container data and the acquired task characteristic data including task workload and weight;
and S6, after the optimal resource allocation scheme is obtained through the fair efficiency balancing method under the two conditions, the cloud rendering resource management system performs task division processing and container scheduling according to the allocation scheme, improves the parallel processing efficiency of the tasks by utilizing available resources to the maximum extent under the condition of ensuring basic fair allocation, and returns the result to the user in time after the tasks are completed.
Further, in order to better implement the present invention, in S4, based on the acquired available cloud container data and the acquired task characteristic data, including the task workload and the weight, a fair efficiency balancing method under resource non-competition is adopted to analyze an optimal resource allocation scheme, and a general generation process is as follows:
s41, uniformly representing sample data based on available cloud container data acquired by the cloud rendering resource management system and acquired newly-added task characteristic data including weight and workload;
s42, analyzing the expected container request quantity of all the newly added tasks based on the obtained weight data of the newly added tasks and the computing capacity of the cloud container, and using the expected container request quantity as important parameter data for guiding fair distribution;
s43, based on the acquired task resource demand data and weight, adopting a proposed weighted acceleration ratio fairness distribution principle, namely according to the principle that the execution speed of the tasks is proportional to the weight of the tasks, carrying out first round of container resource distribution based on a constraint function of fairness guarantee and efficiency maximization so as to realize fairness distribution among user tasks, guarantee user satisfaction and update the quantity of resources for realizing fairness distribution of the user tasks;
s44, acquiring the residual available container resources in the cloud rendering application based on the quantity of the resources acquired after the first round of fair allocation of the user tasks, allocating the residual containers one by utilizing a greedy strategy-based hyper-heuristic algorithm and a maximum performance index-based objective function, and updating the resource allocation results in real time, so that the available resources are utilized to the maximum extent to improve the execution efficiency of the tasks;
s45, judging whether any idle container is not allocated or not based on the acquired number of available servers and the resource allocation result of S44, if so, returning to S44, and otherwise, entering S46;
s46, based on the resource allocation results of S43 and S44, acquiring the tasks which are not allocated to the container, and increasing the weight of the tasks according to the weight function determined by historical data evaluation;
and S47, updating the weight of the newly added task and the number of the containers distributed by the newly added task, and outputting the updated weight to the resource management system.
Further, in order to better implement the present invention, in S5, based on the acquired available cloud container data and the acquired task feature data, including the task workload and the weight, a general generation process of analyzing the optimal resource allocation scheme by using a fair efficiency trade-off method under resource competition is as follows:
s51, uniformly representing sample data based on available cloud container data acquired by the cloud rendering resource management system and acquired newly-added task characteristic data including weight and workload;
s52, analyzing the expected container request quantity of all newly added tasks based on the acquired weight data of the newly added tasks and the computing capacity of the cloud container, and taking the expected container request quantity as important parameter data for guiding fair distribution;
s53, based on the acquired task resource demand data and weight, the available container resources are not considered, the proposed weighted acceleration ratio fairness distribution principle is adopted, and meanwhile, based on the constraint function of a fairness distribution target, the minimum container quantity meeting the basic fairness guarantee of all the newly-added user tasks is acquired, and virtual resource distribution is carried out;
s54, based on the available cloud container data and the minimum container quantity meeting fairness guarantee obtained in the S53, adopting a greedy strategy-based hyper-heuristic algorithm and a minimum performance index-based objective function, deleting the allocated virtual containers one by one, and updating the container quantity virtually allocated to the user task in real time, so that a concession is made in the aspect of acceleration ratio fairness, and the task execution efficiency is guaranteed to the maximum extent;
s55, judging whether the total number of the containers virtually allocated by all the tasks is the same as the number of the available containers applied, if so, entering S53, otherwise, returning to S54;
s56, based on the resource allocation results of S53 and S54, acquiring tasks which are not allocated to the container, evaluating a determined weight function according to historical data, and increasing the weight of the tasks;
and S57, updating the weight of the newly added task and the number of the containers distributed by the task, and outputting the weight and the number of the containers to the cloud rendering resource management system.
The beneficial effects of the invention are:
1. the problems that the QoS requirement of task diversification is difficult to guarantee, rendering efficiency is low, user satisfaction is low and the like are caused by the fact that the characteristics of diversity of cloud user requirements, parallelism of rendering tasks, scale of cloud resources and the like are rarely considered in the conventional fair distribution principle, so that the fair distribution principle based on the weighted acceleration ratio is formulated by fully utilizing the characteristics of parallelism of rendering tasks and the resource standardization of cloud containers, and the central idea is that the acceleration of task execution time is in direct proportion to the weight determined by task remuneration and workload. The fair allocation principle can realize the QoS requirements of different users, and is superior to a Max-min fair resource allocation method in the aspect of realizing the user satisfaction degree.
2. Due to the dynamics of the cloud environment, the randomness of the task arrival and the complexity of the parallel features of the rendering tasks, the number of available cloud containers can not completely meet the resource requirements of all the tasks under general conditions, and therefore the fairness can not be perfectly guaranteed.
3. Efficiency and fairness measurement are two important indexes of a task scheduling method in the evaluation cloud service, however, the fairness resource scheduling method based on user satisfaction constraint can reduce the task execution efficiency of a system under most conditions, so the method provides fairness measurement indexes and acceleration ratio total share measurement indexes through the analysis of user requirements and constraint functions, and establishes a fairness relaxation strategy by adopting multi-index constraint and a greedy strategy-based hyper-heuristic algorithm.
Drawings
FIG. 1 is a schematic diagram of fair distribution of a rendering service in a cloud environment according to the present invention, where P represents reward, and L represents workload;
FIG. 2 is a general flowchart of a method for cloud rendering container resource scheduling based on fairness and efficiency tradeoffs in accordance with the present invention;
FIG. 3 is a flowchart illustrating the updating of the cloud container scheduling scheme under the resource non-contention condition according to the present invention;
fig. 4 is a flowchart illustrating updating of a cloud container scheduling scheme under resource contention according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. It should be apparent that the described embodiments are only some of the embodiments of the present invention, and not all of the embodiments. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention.
In the description of the present invention, it should be noted that the terms "disposed," "connected," and "connected" are to be construed broadly and may be, for example, fixedly connected, detachably connected, or integrally connected unless otherwise explicitly stated or limited. Either mechanically or electrically. They may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood in specific cases to those skilled in the art.
Fig. 1 to 4 show an embodiment of the present invention, in which a highly realistic parallel rendering system established based on the arilocos is used as a background implementation platform, and a data acquisition link is log information of the system, including input/output information, user information, and the like.
Referring to fig. 2, a specific implementation process of the cloud rendering container resource scheduling method based on fairness and efficiency tradeoff according to the present invention includes the following steps:
A. data such as rendering task feature types, user remuneration, submission time and the like are collected through a cloud rendering resource management system, and based on the collected data, task workload and weighted values are analyzed, and all task requirements are summarized:
the weight w correlation calculation process is as follows:
where subscripts A and B represent task A and task B, P represents reward, and L represents task workload. Thus, the total demand of a task can be represented by the sum of the weights: omega = ∑ Σ k w k (k ∈ κ = {1,2, \8230;, n }), κ being the task set.
B. Collecting available cloud containers for rendering services through a cloud rendering resource management system, analyzing available container computing capacity, and summarizing the number c of all available cloud containers t ；
C. Judging whether the existing resources can meet the task computing requirements of all users or not based on the acquired task total requirements and the computing capacity of the existing available cloud container, namely if (c) t ≥ω)；
D. If the judgment result of the step C is yes, analyzing the optimal resource allocation scheme by adopting a fair efficiency balancing method under the resource non-competition condition based on the acquired available cloud container data and the acquired task characteristic data including the task workload and the weight;
D1. uniformly expressing sample data based on available cloud container data acquired by a cloud rendering resource management system and acquired newly added task characteristic data including newly added task number k, weight w and workload l, wherein the workload l is the execution time of a rendering task single thread under the current container configuration;
D2. weight data and cloud container calculation energy based on acquired newly added tasksAnalyzing the expected quantity omega = ∑ Σ of the container requests of all newly added tasks k w k As important parameter data for guiding fair distribution;
D3. based on the acquired task resource demand data and weight, a weighted acceleration ratio fairness allocation principle is adopted, namely, according to the principle that the execution speed of the tasks is proportional to the weight of the tasks, a multi-target constraint function is used for carrying out first round of container resource allocation, and each task is allocated
A container and update the weight of any task k to w k ′＝r k To implement fair distribution among user tasks and update the quantity of resources for implementing fair distribution of user tasks;
D4. obtaining the residual available container resource c in the cloud service based on the resource quantity obtained after the first round of fair allocation of the user task t -sum∑ k r k And based on the objective function of the total share measure of the acceleration ratio
And performing multi-round resource allocation by using the remaining containers one by one with a greedy strategy-based hyper-heuristic algorithm, S k The estimation process of (c) is as follows:
according to the parallelism-prone feature of the rendering task, the expected execution time of task k may be calculated as:
by t k Representing the expected execution time of task k, by r k (r k A positive integer) represents the number of servers assigned to task k. Knowing each rendering task
Will be assigned a weight w k Acceleration value of its execution time of
The efficiency of execution can be expressed as
The shorter the execution time, the higher the execution efficiency. Then, an equivalent of the fairness definition formula can be derived:
wherein, S is a positive variable, called the total share measurement of acceleration ratio (TS-AR), for short. The invention aims to distribute the servers fairly under the application of the existing resource capacity to reach the upper limit of S, thereby maximizing the execution efficiency of the cloud application. How to achieve the maximum TS-AR based on weighted acceleration-ratio fairness is explained next.
As shown in the formula (2),
it can therefore be further deduced that:
then, the equivalent of equation (4) can be derived, calculating the number of servers that should be assigned to task k:
due to r k Is a positive integer, while the number of available servers is limited when rendering tasks are submitted to the application. Thus, the sum of all servers that can be allocated should not exceed the application's existing cloud resource capacity:
based on the above constraints, the following solves the problem of maximizing execution efficiency while maintaining weighted acceleration ratio fairness:
the solution for S can be derived based on the constraints:
according to inequality (8), the maximum TS-AR index S max Can be expressed as:
assuming for each task
More than one container is allocated, and S is calculated k Then comparing to obtain the maximum S, and distributing a container to the task m with the maximum S, and updating the distribution result of the task m to be r m ＝r m +1 and weight w m ′＝w m ' +1 and the remaining available resources c t ＝c t -1；
D5. Based on the obtained number of available containers and the resource allocation result of D4, whether any free container is unallocated or not is judged, namely, whether c t >0, if yes, returning to the step D4, otherwise, entering the step D6;
D6. based on the resource allocation results of the steps D3 and D4, tasks which are not allocated to the container are obtained, a weight function epsilon determined according to historical data evaluation is used for increasing the weight w of the task k k ＝ε*w k ′；
D7. Updating the weight w of the newly added task k And the number r of containers it dispenses k And output to the resource management system;
E. if the judgment result of the step C is negative, analyzing the optimal resource allocation scheme by adopting a fair efficiency balancing method under resource competition based on the acquired available cloud container data and the acquired task characteristic data including task workload and weight;
E1. uniformly representing sample data based on available cloud container data acquired by a cloud rendering resource management system and acquired newly-added task characteristic data including weight w and workload l;
E2. weight data w based on acquired newly added tasks k And computing power of cloud container c t Analyzing the expected container request quantity omega = ∑ Σ of all newly added tasks k w k As important parameter data for guiding fair distribution;
E3. based on the acquired task resource demand data and weight, the available container resources are not considered, a weighted acceleration ratio fairness distribution principle is adopted, and based on a constraint function of a fairness distribution target, the minimum container quantity c meeting the basic fairness guarantee of all newly-added user tasks is acquired t ' = omega, and virtual resource allocation is performed, the number of containers virtually allocated for the task k is r k ＝w k 。
E4. Based on the available cloud container data and the minimum container number meeting the fairness guarantee obtained by E3, deleting the allocated containers one by adopting a greedy strategy-based hyper-heuristic algorithm and a constraint function of a target S of a minimum performance index. Suppose that one container assigned to task k is reduced and S in this case is calculated k . Then the maximum S is obtained k And reducing one container r of the task k having the maximum S value k ＝r k -1, updating the number of containers r planned to be allocated to the user task k And the total number of containers c planned to be allocated to all tasks t ′＝c t ' -1. In container allocation, the invention selects the allocation scheme with the largest S so as to improve the execution efficiency.
E5. Determining the total number of containers c planned to be allocated to all tasks t ' whether or not the same number of usable containers as the application c t If yes, entering the step E3, otherwise, returning to the step E4;
E6. based on the resource allocation results of the steps E3 and E4, acquiring the tasks which are not allocated to the container, evaluating the determined weight function epsilon according to the historical data, and increasing the weight w of the tasks k ＝ε*w k ′；
E7. Updating the weight w of the newly added task k And the number r of containers it dispenses k And output to the resource management system.
F. After the optimal resource allocation scheme is obtained through a fair efficiency balancing method under two conditions, the cloud rendering resource management system performs resource allocation on the tasks according to the allocation scheme, improves the parallel processing efficiency of the tasks by utilizing available resources to the maximum extent under the condition of ensuring basic fair allocation, and returns the result to the user in time after the tasks are completed.
Finally, the above embodiments are only used for illustrating the technical solutions of the present invention and are not limited, and other modifications or equivalent substitutions made by the technical solutions of the present invention by the persons skilled in the art should be covered within the scope of the claims of the present invention as long as they do not depart from the spirit and scope of the technical solutions of the present invention.
Claims (3)
1. A cloud rendering container resource scheduling method based on fairness and efficiency trade-off is characterized by comprising the following steps:
s1, deploying rendering services based on a cloud native technology, collecting rendering task feature types, user requirements and submission time data through a cloud rendering resource management system, analyzing task workloads and weighted values by using the collected data, and summarizing all task requirements;
s2, collecting available cloud containers of the rendering service through a cloud rendering resource management system, analyzing the computing capacity of the available containers, and summarizing all the available cloud containers;
s3, judging whether the existing resources can meet the task computing requirements of all users or not based on the acquired total task requirements and the computing capacity of the existing available cloud container;
s4, if the judgment result of the S3 is positive, determining an optimal resource allocation scheme by adopting a fair efficiency balancing method under resource non-competition based on the acquired available cloud container data and the acquired task characteristic data including task workload and weight;
s5, if the judgment result of the S3 is negative, determining an optimal resource allocation scheme by adopting a fair efficiency balancing method under resource competition based on the acquired available cloud container data and the acquired task characteristic data including task workload and weight;
and S6, after the optimal resource allocation scheme is obtained through the fair efficiency balancing method under the two conditions, the cloud rendering resource management system performs task division processing and container scheduling according to the allocation scheme, improves the parallel processing efficiency of the tasks by utilizing available resources to the maximum extent under the condition of ensuring basic fair allocation, and returns the result to the user in time after the tasks are completed.
2. The fair and efficiency trade-off based cloud rendering container resource scheduling method of claim 1, wherein:
in the S4, based on the acquired available cloud container data and the acquired task characteristic data, including the task workload and the weight, the optimal resource allocation scheme is analyzed by using a fair efficiency balance method under resource noncompetitive conditions, and a general generation process is as follows:
s41, uniformly representing sample data based on available cloud container data acquired by the cloud rendering resource management system and acquired newly-added task characteristic data including weight and workload;
s42, analyzing the expected container request quantity of all the newly added tasks based on the obtained weight data of the newly added tasks and the computing capacity of the cloud container, and using the expected container request quantity as important parameter data for guiding fair distribution;
s43, based on the acquired task resource demand data and weight, adopting a proposed weighted acceleration ratio fairness distribution principle, namely according to the principle that the execution speed of the tasks is proportional to the weight of the tasks, carrying out first round of container resource distribution based on a constraint function of fairness guarantee and efficiency maximization so as to realize fairness distribution among user tasks, guarantee user satisfaction and update the quantity of resources for realizing fairness distribution of the user tasks;
s44, acquiring the residual available container resources in the cloud rendering application based on the quantity of the resources acquired after the first round of fair allocation of the user tasks, allocating the residual containers one by utilizing a greedy strategy-based hyper-heuristic algorithm and a maximum performance index-based objective function, and updating the resource allocation results in real time, so that the available resources are utilized to the maximum extent to improve the execution efficiency of the tasks;
s45, judging whether any idle container is not allocated or not based on the acquired number of available servers and the resource allocation result of S44, if so, returning to S44, and otherwise, entering S46;
s46, based on the resource allocation results of S43 and S44, acquiring the tasks which are not allocated to the container, and increasing the weight of the tasks according to the weight function determined by historical data evaluation;
and S47, updating the weight of the newly added task and the number of the containers distributed by the newly added task, and outputting the updated weight to the resource management system.
3. The fair and efficiency trade-off based cloud rendering container resource scheduling method of claim 1, wherein:
in S5, based on the acquired available cloud container data and the acquired task feature data, including the task workload and the weight, a general generation process of analyzing the optimal resource allocation scheme by using a fair efficiency tradeoff method under resource competition is as follows:
s51, uniformly representing sample data based on available cloud container data acquired by the cloud rendering resource management system and acquired newly-added task characteristic data including weight and workload;
s52, analyzing the expected container request quantity of all newly added tasks based on the acquired weight data of the newly added tasks and the computing capacity of the cloud container, and taking the expected container request quantity as important parameter data for guiding fair distribution;
s53, based on the acquired task resource demand data and weight, the available container resources are not considered, the proposed weighted acceleration ratio fairness distribution principle is adopted, and meanwhile, based on the constraint function of a fairness distribution target, the minimum container quantity meeting the basic fairness guarantee of all the newly-added user tasks is acquired, and virtual resource distribution is carried out;
s54, based on the available cloud container data and the minimum container quantity meeting fairness guarantee obtained in the S53, adopting a greedy strategy-based hyper-heuristic algorithm and a minimum performance index-based objective function, deleting the allocated virtual containers one by one, and updating the container quantity virtually allocated to the user task in real time, so that a concession is made in the aspect of acceleration ratio fairness, and the task execution efficiency is guaranteed to the maximum extent;
s55, judging whether the total number of the containers virtually allocated by all the tasks is the same as the number of the available containers applied, if so, entering S53, otherwise, returning to S54;
s56, acquiring tasks which are not allocated to the container based on the resource allocation results of S53 and S54, and increasing the weight of the tasks according to a weight function determined by historical data evaluation;
and S57, updating the weight of the newly added task and the number of the containers distributed by the newly added task, and outputting the updated weight and the number of the containers to the cloud rendering resource management system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210872619.6A CN115333962A (en) | 2022-07-20 | 2022-07-20 | Cloud rendering container resource scheduling method based on fairness and efficiency balance |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210872619.6A CN115333962A (en) | 2022-07-20 | 2022-07-20 | Cloud rendering container resource scheduling method based on fairness and efficiency balance |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115333962A true CN115333962A (en) | 2022-11-11 |
Family
ID=83919460
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210872619.6A Pending CN115333962A (en) | 2022-07-20 | 2022-07-20 | Cloud rendering container resource scheduling method based on fairness and efficiency balance |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115333962A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103268253A (en) * | 2012-02-24 | 2013-08-28 | 苏州蓝海彤翔系统科技有限公司 | Intelligent scheduling management method for multi-scale parallel rendering jobs |
| CN205193879U (en) * | 2015-10-20 | 2016-04-27 | 国家超级计算深圳中心(深圳云计算中心) | Cloud calculates system of playing up |
| CN107193658A (en) * | 2017-05-25 | 2017-09-22 | 重庆工程学院 | Cloud computing resource scheduling method based on game theory |
| US20190140918A1 (en) * | 2017-11-09 | 2019-05-09 | Futurewei Technologies, Inc. | Method and system for provisioning resources in cloud computing |
| US20210349995A1 (en) * | 2018-09-17 | 2021-11-11 | Georgia Tech Research Corporation | Systems and Methods for Protecting Cache and Main-Memory from Flush-Based Attacks |
-
2022
- 2022-07-20 CN CN202210872619.6A patent/CN115333962A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103268253A (en) * | 2012-02-24 | 2013-08-28 | 苏州蓝海彤翔系统科技有限公司 | Intelligent scheduling management method for multi-scale parallel rendering jobs |
| CN205193879U (en) * | 2015-10-20 | 2016-04-27 | 国家超级计算深圳中心(深圳云计算中心) | Cloud calculates system of playing up |
| CN107193658A (en) * | 2017-05-25 | 2017-09-22 | 重庆工程学院 | Cloud computing resource scheduling method based on game theory |
| US20190140918A1 (en) * | 2017-11-09 | 2019-05-09 | Futurewei Technologies, Inc. | Method and system for provisioning resources in cloud computing |
| US20210349995A1 (en) * | 2018-09-17 | 2021-11-11 | Georgia Tech Research Corporation | Systems and Methods for Protecting Cache and Main-Memory from Flush-Based Attacks |
Non-Patent Citations (2)
| Title |
|---|
| 崔嘉;: "云计算环境下浮点和整数同时多线程资源配置方法研究" * |
| 李秀林: "SaaS环境下易并行应用的性能分析与优化研究" * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Toka et al. | Machine learning-based scaling management for kubernetes edge clusters | |
| CN111953758B (en) | Edge network computing unloading and task migration method and device | |
| CN110489229B (en) | Multi-target task scheduling method and system | |
| CN110287245B (en) | Method and system for scheduling and executing distributed ETL (extract transform load) tasks | |
| CN108419274B (en) | Heterogeneous wireless network selection method based on utility function | |
| CN108092804B (en) | Q-learning-based power communication network utility maximization resource allocation strategy generation method | |
| CN111124689A (en) | Dynamic allocation method for container resources in cluster | |
| CN111641973A (en) | Load balancing method based on fog node cooperation in fog computing network | |
| CN111614754B (en) | Fog-calculation-oriented cost-efficiency optimized dynamic self-adaptive task scheduling method | |
| US8813087B2 (en) | Managing a workload in a cluster of computing systems with multi-type operational resources | |
| CN113822456A (en) | Service combination optimization deployment method based on deep reinforcement learning in cloud and mist mixed environment | |
| CN115480876A (en) | Cloud computing task scheduling method and system based on ant colony algorithm optimization | |
| CN112559135A (en) | QoS-based container cloud resource scheduling method | |
| CN113596868A (en) | 5G network slice resource management mechanism based on SDN and NFV | |
| CN117155942A (en) | Micro-service dynamic self-adaptive client load balancing method and system | |
| More et al. | Energy-aware VM migration using dragonfly–crow optimization and support vector regression model in Cloud | |
| CN108650191B (en) | Decision method for mapping strategy in virtual network | |
| CN112130927B (en) | Reliability-enhanced mobile edge computing task unloading method | |
| CN113139639A (en) | MOMBI-based smart city application-oriented multi-target calculation migration method and device | |
| CN112882805A (en) | Profit optimization scheduling method based on task resource constraint | |
| CN115333962A (en) | Cloud rendering container resource scheduling method based on fairness and efficiency balance | |
| CN115361453B (en) | Load fair unloading and migration method for edge service network | |
| Bensalem et al. | Towards optimal serverless function scaling in edge computing network | |
| CN106447302A (en) | Intelligent order distributing method | |
| CN115016889A (en) | Virtual machine optimization scheduling method for cloud computing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |