CN112363827A - Multi-resource index Kubernetes scheduling method based on delay factors - Google Patents

Multi-resource index Kubernetes scheduling method based on delay factors Download PDF

Info

Publication number
CN112363827A
CN112363827A CN202011162367.5A CN202011162367A CN112363827A CN 112363827 A CN112363827 A CN 112363827A CN 202011162367 A CN202011162367 A CN 202011162367A CN 112363827 A CN112363827 A CN 112363827A
Authority
CN
China
Prior art keywords
cloud computing
computing task
resource
delay
scheduling
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
Application number
CN202011162367.5A
Other languages
Chinese (zh)
Inventor
李克文
吴雪锋
张达
杨建涛
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China University of Petroleum East China
Original Assignee
China University of Petroleum East China
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by China University of Petroleum East China filed Critical China University of Petroleum East China
Priority to CN202011162367.5A priority Critical patent/CN112363827A/en
Publication of CN112363827A publication Critical patent/CN112363827A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5083—Techniques for rebalancing the load in a distributed system
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
    • G06F9/5038—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals considering the execution order of a plurality of tasks, e.g. taking priority or time dependency constraints into consideration
    • 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
    • H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
    • H04L67/1004—Server selection for load balancing
    • H04L67/1008—Server selection for load balancing based on parameters of servers, e.g. available memory or workload
    • 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
    • H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
    • H04L67/1004—Server selection for load balancing
    • H04L67/101—Server selection for load balancing based on network conditions
    • 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
    • H04L67/50—Network services
    • H04L67/60—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources
    • H04L67/61—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources taking into account QoS or priority requirements

Landscapes

  • Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Hardware Design (AREA)
  • Multi Processors (AREA)

Abstract

The invention discloses a multi-resource index Kubernets scheduling method based on delay factors, which is characterized in that the priority of a cloud computing task is evaluated according to a resource request of the cloud computing task and a specific evaluation criterion; setting scheduling delay for the cloud computing task according to the cloud computing task priority and the computing resource state; inserting the cloud computing task into a delay sequence according to the scheduling delay of the cloud computing task; and performing multi-resource index Node matching of dynamic weight according to the resource occupancy rate of the cloud computing task and the Node resource idle state, and distributing the cloud computing task to the optimal Node according to the matching degree. According to the cloud computing resource scheduling method and device, efficient resource scheduling of the cloud computing tasks for different types of resources is achieved through multi-resource index Node matching of the delay factors and the dynamic weights, timely response of the cloud computing tasks is guaranteed, the utilization rate of the cloud computing resources is improved, and load balancing among the resources is achieved.

Description

Multi-resource index Kubernetes scheduling method based on delay factors
Technical Field
The invention belongs to the technical field of cloud computing, and particularly relates to a multi-resource index Kubernetes scheduling method based on a delay factor.
Background
The cloud computing technology is the trend under the current wave and has huge market prospect. Cloud computing is going deep into every corner of life, leading to a thorough revolution in various industries. Kubernets is a container cloud arrangement management system which is most widely applied in the field of cloud computing, and is an important support technology of cloud computing services.
The traditional Kubernetes scheduling method carries out Node scheduling according to two resource indexes of a Node memory and a CPU, resources related to a cloud computing task comprise the CPU and the memory, and also comprise bandwidth, a magnetic disc and a GPU, and the traditional scheduling method cannot meet the resource scheduling requirement of multiple resource indexes in a cloud computing scene. In addition, the resource request response of the cloud computing task is time-ordered, while the traditional Kubernetes scheduling belongs to static resource scheduling, and resource requests in a period of time in the future are not fully considered, so that the problems that the computing task cannot be responded in time, the resource utilization rate is low, the resource load is unbalanced and the like can be caused.
Disclosure of Invention
In order to overcome the defects of the traditional Kubernetes scheduling method, the invention provides a multi-resource index Kubernetes scheduling method based on delay factors, which fully considers resource requests in a period of time in the future, sets scheduling delay for computing tasks according to task priorities and computing resource states, carries out multi-resource index Node matching of dynamic weights aiming at cloud computing tasks of various resource indexes, realizes efficient resource scheduling of the cloud computing tasks, ensures timely response of the cloud computing tasks, improves the utilization rate of the cloud computing resources and realizes load balance among the resources.
In order to achieve the purpose, the technical scheme of the invention is as follows:
A. evaluating the priority of the cloud computing task: calculating the required resource ratio s according to the CPU, memory, bandwidth, disk and GPU resource requests of the cloud computing taskiAccording to siPriority p of cloud computing task according to specific evaluation criterioniEvaluation was carried out.
B. And (3) cloud computing task scheduling delay setting: and setting scheduling delay for the cloud computing task according to the cloud computing task priority and the computing resource state. Longer scheduling delays are set for low priority tasks and shorter or no delays are set for high priority tasks. Longer scheduling delays are set when computing resources are tight, and shorter or no delays are set when computing resources are abundant.
C. Cloud computing task delay: and inserting the cloud computing task into the delay sequence according to the scheduling delay of the cloud computing task, and waiting for the delay to finish the scheduling of the cloud computing task Node.
D. Scheduling a cloud computing task Node: node preselection is carried out according to the cloud computing task resource request, and nodes which do not meet the requirement are excluded; performing dynamic-weight multi-resource index Node matching in the alternative nodes, wherein the resource occupancy rate of a cloud computing task determines the dynamic weight size of a matching function, and the idle state of each resource of the nodes determines the matching degree; and distributing the cloud computing task to the optimal Node according to the matching degree, and if the distribution fails, re-entering a delay queue.
The invention has the beneficial effects that: resource requests in a period of time in the future are fully considered, computing task delay scheduling according to task priorities and computing resource states is achieved, dynamic-weight multi-resource-index Node matching is achieved for cloud computing tasks with multiple resource indexes, efficient resource scheduling of the cloud computing tasks is achieved, timely response of the cloud computing tasks is guaranteed, the utilization rate of the cloud computing resources is improved, and load balancing among resources is achieved.
Drawings
FIG. 1 is a model architecture diagram of the present invention
Detailed Description
A. Evaluating the priority of the cloud computing task: and evaluating the priority of the cloud computing task according to the specific resource request condition of the cloud computing task. Calculating the required resource ratio s according to the resource request of the cloud computing taski，siThe calculation formula is as follows:
si＝(scpu+smem+sbandwidth+sdisk+sgpu)/5，
Figure BDA0002744765430000021
Figure BDA0002744765430000031
Figure BDA0002744765430000032
Figure BDA0002744765430000033
Figure BDA0002744765430000034
wherein, the cpui、memi、bandwidthi、diski、gpuiRespectively representing the computing resources requested by the cloud computing task i. CPU (Central processing Unit)average、memaverage、bandwidthaverage、diskaverage、gpuaverageRespectively representing the average of the computing resources of all nodes. scpu、smem、sbandwidth、sdisk、sgpuRespectively representing the respective resource occupation ratios of the computing resources required by the cloud computing task i.
According to siPriority p of cloud computing task according to specific evaluation criterioniAnd (4) evaluating, wherein evaluation criteria are as follows: siWhen < 5%, piIs 1; s is more than or equal to 5%iWhen < 20%, piIs 2; s is more than or equal to 20%iWhen < 35%, piIs 3; s is more than or equal to 35%iWhen < 50%, piIs 4; siWhen p is more than or equal to 50 percentiIs 5; p is a radical ofiThe larger the task priority.
B. And (3) cloud computing task scheduling delay setting: due to the time sequence of the cloud computing task request, the following two situations are considered: r1For a cloud computing resource task request with priority 3, R2For a cloud computing resource task request with priority 4,
(1) for R in case of shortage of computing resources1、R2Carry out Node matching and satisfy R2The matched Node of the requirement is the Node1Satisfy R1The matched Node of the requirement is the Node1、Node2Corresponding to a degree of matching of m1、m2(m1＞m2). If T is at time R1When coming, the system distributes it to Node1Then T + T time R2Coming, Node1Resource deficiency, R2No response is available.
(2) In the same case for R1、R2Conducting Node matching, R1、R2All the optimally matched nodes are nodes1Corresponding to a degree of matching of m3、m4(m3＜m4)，R1、R2All the sub-optimal matching nodes are nodes2Corresponding to a degree of matching of m5、m6(m5＞m6). If T is at time R1When coming, the system distributes it to Node1Time T + T R2When coming, the system distributes it to Node2。
Obviously, the traditional allocation scheme has various defects in the time sequence of cloud computing, and cannot ensure the timely response of cloud computing tasks, and cannot meet the efficient utilization of resources of service clusters and the load balance among the resources. In order to avoid the problems, scheduling delay is set for the cloud computing task according to the cloud computing task priority and the computing resource state. Longer scheduling delays are set for low priority tasks and shorter or no delays are set for high priority tasks. Longer scheduling delays are set when computing resources are tight, and shorter or no delays are set when computing resources are abundant. The delay calculation formula is as follows:
Figure BDA0002744765430000041
Figure BDA0002744765430000042
wherein t isiThe delay set for task i, t is one unit time, q is the total resource usage,
Figure BDA0002744765430000043
the total occupation ratio of each computing resource.
C. Cloud computing task delay: and inserting the cloud computing task into the delay sequence according to the scheduling delay of the cloud computing task, and waiting for the delay to finish the scheduling of the cloud computing task Node.
D. Scheduling a cloud computing task Node: node preselection is carried out according to the cloud computing task resource request, and nodes which do not meet the requirement are excluded; and performing dynamic-weight multi-resource index Node matching in the alternative nodes, wherein the resource occupancy rate of the cloud computing task determines the dynamic weight of a matching function, and the idle state of each resource of the nodes determines the matching degree. The matching function calculation formula is as follows:
Figure BDA0002744765430000044
wherein, wcpu、wmem、wbandwidth、wdisk、wgpuIs a dynamic weight, w, of five resourcescpu：wmem：wbandwidth：wdisk：wgpu＝scpu：smem：sbandwidth：sdisk：sgpuAnd is
Figure BDA0002744765430000045
Figure BDA0002744765430000046
Is the idle ratio of each computing resource in the Node.
And distributing the cloud computing task to the optimal Node according to the matching degree, and if the distribution fails, re-entering a delay queue.
The foregoing is only a preferred embodiment of this invention and any person skilled in the art may use the above-described solutions to modify or change the same into equivalent embodiments with equivalent variations. Any simple modification, change or amendment to the above-mentioned embodiments according to the technical solutions of the present invention without departing from the technical solutions of the present invention belong to the protection scope of the technical solutions of the present invention.

Claims (1)

1. A multi-resource index Kubernets scheduling method based on delay factors is characterized by comprising the following steps:
A. evaluating the priority of the cloud computing task:
calculating the required resource ratio s according to the CPU, memory, bandwidth, disk and GPU resource requests of the cloud computing taskiAccording to siPriority p of cloud computing task according to specific evaluation criterioniCarrying out evaluation;
B. and (3) cloud computing task scheduling delay setting:
and setting scheduling delay for the cloud computing task according to the cloud computing task priority and the computing resource state. Longer scheduling delays are set for low priority tasks and shorter or no delays are set for high priority tasks. Setting longer scheduling delay when the computing resources are short, and setting shorter delay or not when the computing resources are abundant;
C. cloud computing task delay:
inserting the cloud computing task into a delay sequence according to the scheduling delay of the cloud computing task, and scheduling the cloud computing task Node after the delay is finished;
D. scheduling a cloud computing task Node:
node preselection is carried out according to the cloud computing task resource request, and nodes which do not meet the requirement are excluded; performing dynamic-weight multi-resource index Node matching in the alternative nodes, wherein the resource occupancy rate of a cloud computing task determines the dynamic weight size of a matching function, and the idle state of each resource of the nodes determines the matching degree; and distributing the cloud computing task to the optimal Node according to the matching degree, and if the distribution fails, re-entering a delay queue.
CN202011162367.5A 2020-10-27 2020-10-27 Multi-resource index Kubernetes scheduling method based on delay factors Pending CN112363827A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011162367.5A CN112363827A (en) 2020-10-27 2020-10-27 Multi-resource index Kubernetes scheduling method based on delay factors

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011162367.5A CN112363827A (en) 2020-10-27 2020-10-27 Multi-resource index Kubernetes scheduling method based on delay factors

Publications (1)

Publication Number Publication Date
CN112363827A true CN112363827A (en) 2021-02-12

Family

ID=74512245

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011162367.5A Pending CN112363827A (en) 2020-10-27 2020-10-27 Multi-resource index Kubernetes scheduling method based on delay factors

Country Status (1)

Country Link
CN (1) CN112363827A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113407347A (en) * 2021-06-30 2021-09-17 北京百度网讯科技有限公司 Resource scheduling method, device, equipment and computer storage medium
CN114640681A (en) * 2022-03-10 2022-06-17 京东科技信息技术有限公司 Data processing method and system
WO2024021489A1 (en) * 2022-07-29 2024-02-01 天翼云科技有限公司 Task scheduling method and apparatus, and kubernetes scheduler

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113407347A (en) * 2021-06-30 2021-09-17 北京百度网讯科技有限公司 Resource scheduling method, device, equipment and computer storage medium
CN113407347B (en) * 2021-06-30 2023-02-24 北京百度网讯科技有限公司 Resource scheduling method, device, equipment and computer storage medium
CN114640681A (en) * 2022-03-10 2022-06-17 京东科技信息技术有限公司 Data processing method and system
CN114640681B (en) * 2022-03-10 2024-05-17 京东科技信息技术有限公司 Data processing method and system
WO2024021489A1 (en) * 2022-07-29 2024-02-01 天翼云科技有限公司 Task scheduling method and apparatus, and kubernetes scheduler

Similar Documents

Publication Publication Date Title
CN112363827A (en) Multi-resource index Kubernetes scheduling method based on delay factors
Li et al. An improved max-min task-scheduling algorithm for elastic cloud
CN108628674A (en) Method for scheduling task, cloud platform based on cloud platform and computer storage media
CN106844051A (en) The loading commissions migration algorithm of optimised power consumption in a kind of edge calculations environment
CN104902001B (en) Web request load-balancing method based on operating system virtualization
CN109582448B (en) Criticality and timeliness oriented edge calculation task scheduling method
CN110308967B (en) Workflow cost-delay optimization task allocation method based on hybrid cloud
CN103793272A (en) Periodical task scheduling method and periodical task scheduling system
CN103699445A (en) Task scheduling method, device and system
CN105007337A (en) Cluster system load balancing method and system thereof
CN108446179B (en) Weight priority Task scheduling method based on load balancing mechanism
CN107295090A (en) A kind of method and apparatus of scheduling of resource
CN103401939A (en) Load balancing method adopting mixing scheduling strategy
CN105471985A (en) Load balance method, cloud platform computing method and cloud platform
CN102110014A (en) Method for balancing loads of virtual machine (VM)
CN109861850A (en) A method of the stateless cloud workflow load balance scheduling based on SLA
Tao et al. Load feedback-based resource scheduling and dynamic migration-based data locality for virtual hadoop clusters in openstack-based clouds
CN102339233A (en) Cloud computing centralized management platform
CN106095581B (en) Network storage virtualization scheduling method under private cloud condition
CN104182278A (en) Method and device for judging busy degree of computer hardware resource
CN105430027A (en) Load balance dynamic pre-allocating method based on a plurality of resource scales
CN112888005A (en) MEC-oriented distributed service scheduling method
CN108984286A (en) A kind of resource regulating method and system of cloud computing platform
CN108388471B (en) Management method based on double-threshold constraint virtual machine migration
CN103176850A (en) Electric system network cluster task allocation method based on load balancing

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