CN114489925A - Containerized service scheduling framework and flexible scheduling algorithm - Google Patents

Containerized service scheduling framework and flexible scheduling algorithm Download PDF

Info

Publication number
CN114489925A
CN114489925A CN202111498294.1A CN202111498294A CN114489925A CN 114489925 A CN114489925 A CN 114489925A CN 202111498294 A CN202111498294 A CN 202111498294A CN 114489925 A CN114489925 A CN 114489925A
Authority
CN
China
Prior art keywords
task
node
scheduling
nodes
algorithm
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
CN202111498294.1A
Other languages
Chinese (zh)
Inventor
曾纪钧
龙震岳
张小陆
梁哲恒
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.)
Guangdong Power Grid Co Ltd
Original Assignee
Guangdong Power Grid Co Ltd
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 Guangdong Power Grid Co Ltd filed Critical Guangdong Power Grid Co Ltd
Priority to CN202111498294.1A priority Critical patent/CN114489925A/en
Publication of CN114489925A publication Critical patent/CN114489925A/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
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors
    • G06F9/45558—Hypervisor-specific management and integration aspects
    • 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
    • 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]
    • G06F9/5061—Partitioning or combining of resources
    • G06F9/5072—Grid computing
    • 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]
    • G06F9/5061—Partitioning or combining of resources
    • G06F9/5077—Logical partitioning of resources; Management or configuration of virtualized resources
    • 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
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors
    • G06F9/45558—Hypervisor-specific management and integration aspects
    • G06F2009/4557—Distribution of virtual machine instances; Migration and load balancing
    • 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
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors
    • G06F9/45558—Hypervisor-specific management and integration aspects
    • G06F2009/45591—Monitoring or debugging support
    • 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
    • G06F9/44—Arrangements for executing specific programs
    • G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
    • G06F9/45533—Hypervisors; Virtual machine monitors
    • G06F9/45558—Hypervisor-specific management and integration aspects
    • G06F2009/45595—Network integration; Enabling network access in virtual machine instances
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2209/00—Indexing scheme relating to G06F9/00
    • G06F2209/50—Indexing scheme relating to G06F9/50
    • G06F2209/502—Proximity

Landscapes

  • Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a containerized service scheduling framework and an elastic scheduling algorithm, which comprises the steps that a user request is subjected to semantic analysis and forwarding through an API gateway of a main node; creating and scheduling the analyzed actual operation objects, and storing all the objects into a key value database; additional controller processes perform health monitoring on the objects in the cluster. The invention integrates a Docker container technology and a Kubenetes container management technology, provides a container technology-based electric power Internet of things edge side service scheduling framework, optimizes and improves initialization, updating, heuristic factor setting and the like of pheromones on the basis of labeling an ant colony algorithm in order to make up the defects of an electric power Internet of things edge side service scheduling algorithm under the container technology-based electric power Internet of things edge side service scheduling framework, and designs an elastic resource dynamic scheduling algorithm based on the ant colony algorithm, so that load balance of edge task allocation is realized, and limited edge resources can exert the maximum effect.

Description

Containerized service scheduling framework and flexible scheduling algorithm
Technical Field
The invention relates to the technical field of power Internet of things, in particular to a containerization service scheduling framework and an elastic scheduling algorithm.
Background
In a new generation of power internet of things, high bandwidth, high density and low time delay are the main characteristics of services. If the original cloud computing mode is still adopted, the speed, the delay and the stability cannot be ensured at all. The edge computing is a computing mode for providing services for users at the edge of a network, and has the characteristics of low time delay and high bandwidth. With the application of the edge computing technology, the service scheduling function is gradually inclined to the edge side. However, the existing power internet of things edge side scheduling strategy has low resource distribution efficiency and poor flexibility, so that the data processing delay is high, and the service expandability is not high.
Disclosure of Invention
This section is for the purpose of summarizing some aspects of embodiments of the invention and to briefly introduce some preferred embodiments. In this section, as well as in the abstract and the title of the invention of this application, simplifications or omissions may be made to avoid obscuring the purpose of the section, the abstract and the title, and such simplifications or omissions are not intended to limit the scope of the invention.
The present invention has been made in view of the above and/or other problems occurring in the existing containerized service scheduling framework and flexible scheduling algorithm.
Therefore, the invention aims to solve the problems that the existing power internet of things edge side scheduling strategy has low resource distribution efficiency and poor flexibility, so that the data processing delay is high and the service expandability is not high.
In order to solve the technical problems, the invention provides the following technical scheme: a containerized service scheduling framework and an elastic scheduling algorithm comprise that a user request is subjected to semantic analysis and forwarding through an API gateway of a main node;
creating and scheduling the analyzed actual operation objects, and storing all the objects into a key value database;
an additional controller process performs health monitoring on objects in the cluster;
based on HPA horizontal dilatation controller in Kubenetes, realize dynamic resource scheduling, the index data that horizontal dilatation controller collected the third party instrument through the index aggregator is gathered to be used for the elasticity scheduling algorithm with the index, calculate the optimal scheme of elasticity resource scheduling, and carry out dynamic dilatation, the specific flow of dilatation:
creating HPA horizontal capacity expansion controller resources;
setting a period of a control manager;
running an application example and collecting the use condition of resources;
tolerances are taken into account.
As a preferred solution of the containerized service scheduling framework and flexible scheduling algorithm of the present invention, wherein: the master node is used as a control center of the whole cluster and mainly responsible for allocation, scheduling and recovery of resources, and all the slave nodes are used as resource pools and are used as division objects of the master node.
As a preferred solution of the containerized service scheduling framework and flexible scheduling algorithm of the present invention, wherein: all nodes adopt virtual machines, the availability of the nodes is guaranteed by the multiple virtual machines, and all the nodes in the whole framework operate as containers through Docker engines.
As a preferred solution of the containerized service scheduling framework and flexible scheduling algorithm of the present invention, wherein: the cluster storage adopts a distributed database etcd, and stores metadata and expected states and current states of all resources in a high-availability key value storage mode.
As a preferred solution of the containerized service scheduling framework and flexible scheduling algorithm of the present invention, wherein: on the basis of the labeled ant colony algorithm, initialization, updating, heuristic factor setting and the like of pheromones are optimized and improved so as to achieve target tasks and meet requirements under the edge computing environment.
The load balance of the edge nodes is measured by adopting the load unbalance degree, the value is taken from 0 to 1, and the smaller the value is, the more uniform the load distribution of the tasks of the edge nodes is, and the higher the overall performance of the system is. The concrete expression is as follows:
Figure RE-GDA0003576882690000021
wherein the content of the first and second substances,
Figure RE-GDA0003576882690000022
respectively represent edgesThe utilization rate of an edge node i (i is 1, …, n) for a CPU, a GPU and a memory;
Figure RE-GDA0003576882690000023
respectively representing the average utilization rate of the whole edge node to CPU, GPU and memory.
As a preferred solution of the containerized service scheduling framework and flexible scheduling algorithm of the present invention, wherein: eta for heuristic factorsijThe main expression is the desired strength, η, of the placement of task i at edge node jijThe larger the value of (c) the greater the probability that the task is placed at the node:
Figure RE-GDA0003576882690000024
wherein Q is an arbitrary constant, and Q is an arbitrary constant,
Figure RE-GDA0003576882690000025
the cosine similarity between the resource required by the task to be distributed and the node idle resource can be defined, which is equivalent to the similarity between the task i and the node j represented by the included angle between the task i and the node j;
the smaller the included angle is, the higher the similarity between the two is represented, and the higher the possibility that the task is allocated at the node is:
Figure RE-GDA0003576882690000031
wherein, A represents the number of the resource types which can be provided by the edge node for the task (three resources of CPU, GPU and memory are used as demand resources);
Figure RE-GDA0003576882690000032
the idle amount of the a-th resource of the edge node i;
Figure RE-GDA0003576882690000033
the demand of the resource a of task j.
As a preferred solution of the containerized service scheduling framework and flexible scheduling algorithm of the present invention, wherein: the distribution between the tasks and the nodes is a mapping relation of many-to-one, the quantity of each resource required by the tasks is ensured to be smaller than the quantity of the idle resources of the nodes in the distribution of the tasks, and the tasks are prevented from being distributed on the nodes with insufficient resources by comparing the demands of the tasks on a CPU (Central processing Unit), a GPU (graphics processing Unit) and a memory with the available resources of the nodes, namely:
Figure RE-GDA0003576882690000034
Figure RE-GDA0003576882690000035
Figure RE-GDA0003576882690000036
wherein the content of the first and second substances,
Figure RE-GDA0003576882690000037
respectively representing the use amounts of a CPU, a GPU and a memory of the task j on the edge node i;
Figure RE-GDA0003576882690000038
respectively representing the available amount of the edge node i relative to the CPU, the GPU and the memory.
As a preferred solution of the containerized service scheduling framework and flexible scheduling algorithm of the present invention, wherein: the pheromone update of the initial ant colony algorithm is volatile for all paths, which results in that the pheromones of the nodes which are not frequently walked are lower and even approach zero. Therefore, the route which is not traveled is not volatilized, and the specific change is as follows:
Figure RE-GDA0003576882690000041
Figure RE-GDA0003576882690000042
where ρ represents the volatilization factor of pheromones and 1- ρ represents the residual factor of pheromones. If Δ τijA value of 0 indicates that the task is not allocated on the node.
As a preferred solution of the containerized service scheduling framework and flexible scheduling algorithm of the present invention, wherein: the overall algorithm flow is as follows:
inputting: the method comprises the following steps of setting an edge node set PM (PM 1, PM2, … and pmn), setting a task requirement set TN (TN 1, TN2, … and tnn), an pheromone heuristic factor alpha, an expected heuristic factor beta, an pheromone volatilization rate rho and other initial requirements.
And (3) outputting: optimal task allocation scheme and load imbalance.
As a preferred solution of the containerized service scheduling framework and flexible scheduling algorithm of the present invention, wherein: initializing the number AntNum of ants and attaching the required resources of the tasks to each task according to the order of task submission;
randomly distributing n ants carrying task requirements on random nodes, and calculating the probability that the kth ant distributes a task i on a node j;
after the kth ant completes all task allocation, locally updating the allocated nodes;
after the ant is deployed, calculating the load imbalance degree of the distribution scheme, comparing the load imbalance degree with a historical record, and recording the optimal distribution scheme and the minimum load imbalance degree;
judging whether all ants are finished or not;
and judging whether the iteration times are met or the preset load balance degree is met, returning the optimal path solution after the algorithm iteration is finished, and otherwise, recalculating the probability that the kth ant distributes the task i to the node j.
The invention has the beneficial effects that: the invention integrates a Docker container technology and a Kubenetes container management technology, provides a container technology-based electric power Internet of things edge side service scheduling framework, optimizes and improves initialization, updating, heuristic factor setting and the like of pheromones on the basis of labeling an ant colony algorithm in order to make up the defects of an electric power Internet of things edge side service scheduling algorithm under the container technology-based electric power Internet of things edge side service scheduling framework, and designs an elastic resource dynamic scheduling algorithm based on the ant colony algorithm, so that load balance of edge task allocation is realized, and limited edge resources can exert the maximum effect.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive exercise. Wherein:
fig. 1 is a scene diagram of a containerized service scheduling framework and an elastic scheduling algorithm.
Detailed Description
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways than those specifically described and will be readily apparent to those of ordinary skill in the art without departing from the spirit of the present invention, and therefore the present invention is not limited to the specific embodiments disclosed below.
Furthermore, reference herein to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one implementation of the invention. The appearances of the phrase "in one embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments.
Example 1
Referring to fig. 1, a first embodiment of the present invention provides a containerized service scheduling framework and an elastic scheduling algorithm, where the containerized service scheduling framework and the elastic scheduling algorithm include that as shown in fig. 1, a distributed master/slave architecture is adopted in an overall architecture, and a master node is used as a control center of a whole cluster and is mainly responsible for resource allocation, scheduling, and recovery. All slave nodes are used as resource pools as partitioning objects of the master node. All the nodes adopt virtual machines, and the multiple virtual machines ensure the availability of the nodes and reduce the possibility of the paralysis of the nodes to the minimum. All nodes in the whole architecture run as containers through a Docker engine. The cluster storage adopts a distributed database etcd, stores metadata and expected states and current states of all resources in a high-availability key value storage mode, and ensures data safety and stable operation of a system.
The whole framework operation flow comprises the following steps: the user requests to perform semantic analysis and forwarding through an API gateway of the main node; creating and scheduling the analyzed actual operation objects, and storing all the objects into a key value database; additional controller processes perform health monitoring on the objects in the cluster.
Based on an HPA horizontal capacity expansion controller in Kubenetes, dynamic resource scheduling is realized. And the horizontal capacity expansion controller aggregates the index data collected by the third-party tool through the index aggregator, so that the index is used for an elastic scheduling algorithm, an optimal scheme for elastic resource scheduling is calculated, and dynamic capacity expansion is performed. Capacity expansion specific process:
HPA level expansion controller resources are created. Setting the number range of the application instances and the CPU average usage limit, setting the resource usage limit of the Pod according to the application type, setting tolerance parameters, and defaulting to 0.1;
the period of the control manager is set. The period is the period of inquiring the resource use condition and is set to be 30 seconds;
and running the application example and collecting the resource use condition. Calculating (sum of index values/set limit) by using the index values and the set values to obtain the number of target adjustment examples;
tolerances are taken into account. If the tolerance range is exceeded, the number of the examples is adjusted, otherwise, the number of the examples of the target adjustment is repeatedly calculated. The number of the target adjusted instances cannot exceed the range of the number of the instances set in the tolerance parameter, and if the number of the target adjusted instances does not exceed the range of the number of the instances set in the tolerance parameter, capacity expansion is carried out; if the target adjustment is exceeded, the capacity is expanded to the maximum number of instances, the index value and the set value are repeatedly used for calculation (the sum of the index values/the set limit), and the number of instances of the target adjustment is obtained.
On the basis of the labeled ant colony algorithm, initialization, updating, heuristic factor setting and the like of pheromones are optimized and improved so as to achieve target tasks and meet requirements under the edge computing environment.
The load balance of the edge nodes is measured by adopting the load unbalance degree, the value is taken from 0 to 1, and the smaller the value is, the more uniform the load distribution of the tasks of the edge nodes is, and the higher the overall performance of the system is. The concrete expression is as follows:
Figure RE-GDA0003576882690000061
this is shown in equation 1, where,
Figure RE-GDA0003576882690000062
respectively representing the utilization rates of the edge nodes i (i is 1, …, n) to a CPU, a GPU and a memory;
Figure RE-GDA0003576882690000063
respectively representing the average utilization rate of the whole edge node to CPU, GPU and memory.
Eta for heuristic factorsijThe main expression is the desired strength, η, of the placement of task i at edge node jijThe larger the value of (c) the greater the probability that the task is placed at the node:
Figure RE-GDA0003576882690000064
this is equation 2, where Q is an arbitrary constant,
Figure RE-GDA0003576882690000071
the cosine similarity between the task waiting to be allocated and the required resource and the node idle resource can be defined, which is equivalent to that the similarity between the task i and the node j is represented by the included angle between the task i and the node j. The smaller the included angle is, the higher the similarity between the two is represented, and the higher the possibility that the task is allocated at the node is:
Figure RE-GDA0003576882690000072
this is formula 3, where a represents the number of resource types that the edge node can provide for the task (three resources, i.e., CPU, GPU, and memory, are required resources);
Figure RE-GDA0003576882690000073
the idle amount of the a-th resource of the edge node i;
Figure RE-GDA0003576882690000074
the demand of the resource a of task j.
From equation 3, it follows: the smaller the matching between the task to be distributed and the edge node, the more the node can adapt to the performance requirement of the task on the node, and the higher the probability that the task is distributed on the node. In the process of task allocation, the matching degree is continuously changed along with the change of time, so that
Figure RE-GDA0003576882690000075
And representing the matching degree of the edge node i and the task j at the moment t, and if the task j is placed on the edge node i, calculating the matching degree of the edge node i and other tasks at the next time and calculating the idle resources after the resources required by the task j are distributed.
The distribution between the tasks and the nodes is a mapping relation of many-to-one, the quantity of each resource required by the tasks is ensured to be smaller than the quantity of the idle resources of the nodes in the distribution of the tasks, and the tasks are prevented from being distributed on the nodes with insufficient resources by comparing the demands of the tasks on a CPU (Central processing Unit), a GPU (graphics processing Unit) and a memory with the available resources of the nodes, namely:
Figure RE-GDA0003576882690000076
Figure RE-GDA0003576882690000077
Figure RE-GDA0003576882690000078
formula 4, formula 5, and formula 6, in that order,
Figure RE-GDA0003576882690000079
respectively representing the use amounts of a CPU, a GPU and a memory of the task j on the edge node i;
Figure RE-GDA00035768826900000710
respectively representing the available amount of the edge node i relative to the CPU, the GPU and the memory.
The pheromone update of the initial ant colony algorithm is volatile for all paths, which results in that the pheromones of the nodes which are not frequently walked are lower and even approach zero. Therefore, the route which is not traveled is not volatilized, and the specific change is as follows:
Figure RE-GDA0003576882690000081
Figure RE-GDA0003576882690000082
equation 7 and equation 8 follow, where ρ represents the volatility factor of the pheromone and 1- ρ represents the retention factor of the pheromone. If Δ τij0 indicates that the task is not allocatedAbove the node.
The overall algorithm flow is as follows:
inputting: the method comprises the following steps of an edge node set PM (PM 1, PM2, … and pmn), a task requirement set TN (TN 1, TN2, … and tnn), an pheromone heuristic factor alpha, an expected heuristic factor beta, an pheromone volatilization rate rho and other initial requirements.
And (3) outputting: optimal task allocation scheme and load imbalance.
Step 1: initializing ant number AntNum and attaching required resources of the tasks to each task according to the order of task submission, iteration times maxIter, heuristic factors alpha and beta and related parameters.
Step 2: and randomly distributing n ants carrying task requirements on random nodes, and calculating the probability that the kth ant distributes the task i on the node j. And then randomly selecting one node from the nodes meeting the conditions as a distribution node of the task by using a roulette mode, and deploying the task on the node. The probability selection formula is:
Figure RE-GDA0003576882690000083
this is the equation 9 in which, among other things,
Figure RE-GDA0003576882690000084
represents the probability of task i in the kth ant selecting node j, tauij(t) pheromone concentration, η, of pathway (i, j)ij(t) represents the heuristic factor of the path (i, j), as in equation (2). allowedkAn allocable list for task i.
And step 3: after the kth ant completes all task assignments, the assigned nodes are locally updated according to equation 5.
And 4, step 4: and after the disposition of the ants is completed, calculating the load imbalance degree of the distribution scheme according to the formula 1, comparing with a historical record, and recording the optimal distribution scheme and the minimum load imbalance degree.
And 5: judging whether all ants are finished or not, and jumping to the step 2 if all ants are not finished; if all ants finish the iteration, calculating and storing a global optimal solution, and finally performing global updating on the pheromone according to an optimal placement scheme and a formula 7.
Step 6: and judging whether the iteration times are met or the preset load balance degree is met, and returning the optimal path solution after the algorithm iteration is finished. Otherwise, returning to the step 2. .
It should be noted that the above-mentioned embodiments are only for illustrating the technical solutions of the present invention and not for limiting, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications or equivalent substitutions may be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, which should be covered by the claims of the present invention.

Claims (10)

1. A containerized service dispatch framework, comprising: comprises the steps of (a) preparing a mixture of a plurality of raw materials,
the user requests to perform semantic analysis and forwarding through an API gateway of the main node;
creating and scheduling the analyzed actual operation objects, and storing all the objects into a key value database;
an additional controller process performs health monitoring on objects in the cluster;
based on HPA horizontal dilatation controller in Kubenetes, realize dynamic resource scheduling, the index data that horizontal dilatation controller collected the third party instrument through the index aggregator is gathered to be used for the elasticity scheduling algorithm with the index, calculate the optimal scheme of elasticity resource scheduling, and carry out dynamic dilatation, the specific flow of dilatation:
creating HPA horizontal capacity expansion controller resources;
setting a period of a control manager;
running an application example and collecting the use condition of resources;
tolerances are taken into account.
2. The containerized services scheduling framework of claim 1, wherein: the master node is used as a control center of the whole cluster and mainly responsible for allocation, scheduling and recovery of resources, and all the slave nodes are used as resource pools and are used as division objects of the master node.
3. The containerized services scheduling framework of claim 2, wherein: all nodes adopt virtual machines, the availability of the nodes is guaranteed by the multiple virtual machines, and all nodes in the whole architecture run as containers through a Docker engine.
4. The containerized services scheduling framework of claim 3, wherein: the cluster storage adopts a distributed database etcd, and stores metadata and expected states and current states of all resources in a high-availability key value storage mode.
5. The flexible scheduling algorithm of the containerized service scheduling framework of any one of claims 1-4, wherein: on the basis of the labeled ant colony algorithm, initialization, updating, heuristic factor setting and the like of pheromones are optimized and improved so as to achieve target tasks and meet requirements under the edge computing environment.
The load balance of the edge nodes is measured by adopting the load unbalance degree, the value is taken from 0 to 1, and the smaller the value is, the more uniform the load distribution of the tasks of the edge nodes is, and the higher the overall performance of the system is. The concrete expression is as follows:
Figure FDA0003401755680000011
wherein the content of the first and second substances,
Figure FDA0003401755680000012
respectively representing the utilization rates of the edge nodes i (i is 1, …, n) to a CPU, a GPU and a memory;
Figure FDA0003401755680000021
respectively representing the average utilization rate of the whole edge node to CPU, GPU and memory.
6. The flexible scheduling algorithm of the containerized services scheduling framework of claim 5, wherein: eta for heuristic factorsijThe main expression is the desired strength, η, of the placement of task i at edge node jijThe larger the value of (c) the greater the probability that the task is placed at the node:
Figure FDA0003401755680000022
wherein Q is an arbitrary constant, and Q is an arbitrary constant,
Figure FDA0003401755680000023
the cosine similarity between the resource required by the task to be distributed and the node idle resource can be defined, which is equivalent to the similarity between the task i and the node j represented by the included angle between the task i and the node j;
the smaller the included angle is, the higher the similarity between the two is represented, and the higher the possibility that the task is allocated at the node is:
Figure FDA0003401755680000024
wherein, A represents the number of the resource types which can be provided by the edge node for the task (three resources of CPU, GPU and memory are used as demand resources);
Figure FDA0003401755680000025
the idle amount of the a-th resource of the edge node i;
Figure FDA0003401755680000026
the demand of the resource a of task j.
7. The flexible scheduling algorithm of the containerized services scheduling framework of claim 6, wherein: the distribution between the tasks and the nodes is a mapping relation of many-to-one, the quantity of each resource required by the tasks is ensured to be smaller than the quantity of the idle resources of the nodes in the distribution of the tasks, and the tasks are prevented from being distributed on the nodes with insufficient resources by comparing the demands of the tasks on a CPU (Central processing Unit), a GPU (graphics processing Unit) and a memory with the available resources of the nodes, namely:
Figure FDA0003401755680000027
Figure FDA0003401755680000028
Figure FDA0003401755680000029
the amount of use of (c);
Figure FDA0003401755680000031
respectively representing the available amount of the edge node i relative to the CPU, the GPU and the memory.
8. The flexible scheduling algorithm of the containerized services scheduling framework of claim 7 wherein: the pheromone update of the initial ant colony algorithm is volatile for all paths, which results in that the pheromones of the nodes which are not frequently walked are lower and even approach zero. Therefore, the route which is not traveled is not volatilized, and the specific change is as follows:
Figure FDA0003401755680000032
Figure FDA0003401755680000033
where ρ represents the volatilization factor of pheromones and 1- ρ represents the residual factor of pheromones. If Δ τijA value of 0 indicates that the task is not allocated on the node.
9. The flexible scheduling algorithm of the containerized services scheduling framework of claim 8, wherein: the overall algorithm flow is as follows:
inputting: the method comprises the following steps of an edge node set PM (PM 1, PM2, … and pmn), a task requirement set TN (TN 1, TN2, … and tnn), an pheromone heuristic factor alpha, an expected heuristic factor beta, an pheromone volatilization rate rho and other initial requirements.
And (3) outputting: optimal task allocation scheme and load imbalance.
10. The flexible scheduling algorithm of the containerized services scheduling framework of claim 9, wherein: initializing the number AntNum of ants and attaching the required resources of the tasks to each task according to the order of task submission;
randomly distributing n ants carrying task requirements on random nodes, and calculating the probability that the kth ant distributes a task i on a node j;
after the kth ant completes all task allocation, locally updating the allocated nodes;
after the ant is deployed, calculating the load imbalance degree of the distribution scheme, comparing the load imbalance degree with a historical record, and recording the optimal distribution scheme and the minimum load imbalance degree;
judging whether all ants are finished or not;
and judging whether the iteration times are met or the preset load balance degree is met, returning the optimal path solution after the algorithm iteration is finished, and otherwise, recalculating the probability that the kth ant distributes the task i to the node j.
CN202111498294.1A 2021-12-09 2021-12-09 Containerized service scheduling framework and flexible scheduling algorithm Pending CN114489925A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111498294.1A CN114489925A (en) 2021-12-09 2021-12-09 Containerized service scheduling framework and flexible scheduling algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111498294.1A CN114489925A (en) 2021-12-09 2021-12-09 Containerized service scheduling framework and flexible scheduling algorithm

Publications (1)

Publication Number Publication Date
CN114489925A true CN114489925A (en) 2022-05-13

Family

ID=81492393

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111498294.1A Pending CN114489925A (en) 2021-12-09 2021-12-09 Containerized service scheduling framework and flexible scheduling algorithm

Country Status (1)

Country Link
CN (1) CN114489925A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114996025A (en) * 2022-08-01 2022-09-02 四川公众项目咨询管理有限公司 Network defense method based on edge technology
CN116450382A (en) * 2023-06-19 2023-07-18 鹏城实验室 Data processing method and system based on function definition
CN116684483A (en) * 2023-08-02 2023-09-01 北京中电普华信息技术有限公司 Method for distributing communication resources of edge internet of things proxy and related products
CN116887357A (en) * 2023-09-08 2023-10-13 山东海博科技信息***股份有限公司 Computing platform management system based on artificial intelligence

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114996025A (en) * 2022-08-01 2022-09-02 四川公众项目咨询管理有限公司 Network defense method based on edge technology
CN116450382A (en) * 2023-06-19 2023-07-18 鹏城实验室 Data processing method and system based on function definition
CN116684483A (en) * 2023-08-02 2023-09-01 北京中电普华信息技术有限公司 Method for distributing communication resources of edge internet of things proxy and related products
CN116684483B (en) * 2023-08-02 2023-09-29 北京中电普华信息技术有限公司 Method for distributing communication resources of edge internet of things proxy and related products
CN116887357A (en) * 2023-09-08 2023-10-13 山东海博科技信息***股份有限公司 Computing platform management system based on artificial intelligence
CN116887357B (en) * 2023-09-08 2023-12-19 山东海博科技信息***股份有限公司 Computing platform management system based on artificial intelligence

Similar Documents

Publication Publication Date Title
CN114489925A (en) Containerized service scheduling framework and flexible scheduling algorithm
CN107273185B (en) Load balancing control method based on virtual machine
CN113193984B (en) Air-space-ground integrated network resource mapping method and system
US8812653B2 (en) Autonomous intelligent workload management
US8380557B2 (en) Multi-tenant database management for service level agreement (SLA) profit maximization
CN110858161A (en) Resource allocation method, device, system, equipment and medium
Deboosere et al. Efficient resource management for virtual desktop cloud computing
CN111163178B (en) Game theory-based service deployment and task unloading method in edge computing
CN109617826A (en) A kind of storm dynamic load balancing method based on cuckoo search
CN113163498B (en) Virtual network resource allocation method and device based on genetic algorithm under 5G network slice
CN111917818B (en) Dynamic matching method for personalized service demands
CN109460301B (en) Method and system for configuring elastic resources of streaming data load
Supreeth et al. Hybrid genetic algorithm and modified-particle swarm optimization algorithm (GA-MPSO) for predicting scheduling virtual machines in educational cloud platforms
KR20180072295A (en) Dynamic job scheduling system and method for supporting real-time stream data processing in distributed in-memory environment
CN108595255B (en) Workflow task scheduling method based on shortest path algorithm in geographically distributed cloud
CN115134371A (en) Scheduling method, system, equipment and medium containing edge network computing resources
Fathi et al. Consolidating VMs in green cloud computing using harmony search algorithm
CN110990160B (en) Static security analysis container cloud elastic telescoping method based on load prediction
CN109067888A (en) The distributed cloudy resource multi-level Fusion management system in strange land
CN111176784A (en) Virtual machine integration method based on extreme learning machine and ant colony system
CN111913800B (en) Resource allocation method for optimizing cost of micro-service in cloud based on L-ACO
Chatterjee et al. A new clustered load balancing approach for distributed systems
Mishra et al. Adaptive scheduling of cloud tasks using ant colony optimization
CN108234617A (en) A kind of resource dynamic dispatching method under the mixing cloud mode towards electric system
CN112437449A (en) Joint resource allocation method and area organizer

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