Micro-service scheduling method and system based on cloud edge collaboration

Abstract

In the micro-service scheduling method and system based on cloud edge collaboration, aiming at an intelligent industrial process scene, a user task request is decomposed into a plurality of micro-services to be scheduled, aiming at each micro-service to be scheduled, edge nodes and / or cloud nodes meeting resource requirements of the micro-service to be scheduled are screened out to construct a candidate node set, and the candidate node set is used for scheduling the micro-service to be scheduled. According to the method, inter-container interference types are subdivided and quantified, an interference index calculation model is constructed for each inter-container interference type, and an objective function is constructed by taking inter-container interference minimization, cluster resource utilization rate improvement and service response time reduction as targets, so that a micro-service scheduling problem is converted into a multi-objective optimization problem; according to the invention, reasonable scheduling of micro-services in an industrial process intelligent monitoring scene is realized, interference between containers is reduced, and service quality of edge applications is improved.

Description

Technical Field

[0001] The present invention belongs to the field of industrial Internet and edge computing technology, and specifically, relates to a microservice scheduling method and system based on cloud-edge collaboration. Background Art

[0002] The intelligent construction of industrial processes refers to the use of new technologies to analyze, integrate and control the resources in the production system, so that the system has the ability to self-operate, self-organize and self-optimize, and make intelligent responses to the behaviors in the system.

[0003] With the gradual advancement of informatization, the construction of intelligent industrial processes cannot be achieved overnight, and the original architecture of early systems cannot adapt to the ever-changing needs. In the intelligent industrial scenario, applications such as license plate recognition, face recognition, and security alarms have high requirements for communication latency, computing speed, and bandwidth. The traditional computing method of collecting data through sensors and uploading it to the cloud platform can no longer meet the existing needs. Therefore, it is of great significance to deploy industrial process application services to the edge of the network and build an intelligent service system at the edge of the industrial process through edge computing to provide users with nearby services and meet the various needs of intelligent industry. However, the resources on the edge side are limited. How to effectively utilize existing infrastructure resources (such as servers, gateways, cameras, etc.), build, uniformly manage and allocate edge resources, and achieve efficient use of edge resources under limited resource conditions is very important.

[0004] At the same time, the demands in industrial processes change frequently, and new applications require rapid deployment and adjustment, which requires managers to provide more agile service building capabilities. With the emergence of microservice architecture, the development and deployment of industrial applications based on microservice architecture can save application development time, and can scale up and down a single microservice, which can improve the flexibility of the system. Before the emergence of container technology, microservices were mainly deployed on virtual machines, and the system overhead was very high. Compared with virtual machines, container technology reduces the virtualization layer of the hardware system, is more lightweight, and can be used as a microservice carrier to achieve rapid deployment of microservices. However, due to the small granularity of containers, container resource management is more complicated. Commonly used Docker scheduling tools, such as Kubernetes and Docker swarm, only provide some simple scheduling strategies and cannot be directly applied to smart industrial scenarios with multiple heterogeneous resources. Therefore, it is very important to design a reasonable microservice scheduling algorithm. Summary of the invention

[0005] This invention focuses on the resource construction and cloud-edge collaborative microservice scheduling problems under intelligent industrial processes, proposes a microservice scheduling method and system based on cloud-edge collaboration, adopts lightweight container technology and open source edge computing framework to construct resources in industrial processes, manages resources through edge service systems to improve infrastructure resource utilization, and converts the microservice scheduling problem into a multi-objective optimization problem with the goals of minimizing interference between containers, improving cluster resource utilization, and reducing service response time. A cloud-edge collaborative microservice scheduling algorithm is proposed for solving it, which reduces interference between containers and improves the service quality of edge applications.

[0006] The present invention is implemented by the following technical solutions:

[0007] A microservice scheduling method based on cloud-edge collaboration is proposed, including:

[0008] Get user task requests and decompose them into multiple microservices to be scheduled;

[0009] For each microservice to be scheduled, select edge nodes and / or cloud nodes that meet its resource requirements and build a candidate node set;

[0010] The objective function is constructed with the goal of minimizing service delay time, minimizing scheduling time, and minimizing interference between containers. The model for minimizing interference between containers is: , To calculate the interference index, is the CPU resource requested by the container, is the CPU resources available to the node; is the network interference index, is the bandwidth consumed by the container, It is the I / O resources available to the node; = is the I / O interference index, are the I / O resources requested by the container, is the I / O resources available to the node;

[0011] Solve the objective function to obtain the optimal microservice scheduling result.

[0012] In some embodiments of the present invention, after decomposing to obtain a plurality of microservices to be scheduled, the method further includes:

[0013] Generate a microservice task graph, where nodes represent microservices and edges represent dependencies between microservices;

[0014] For each microservice to be scheduled, minimize cross-cluster or cross-host calls between microservices.

[0015] In some embodiments of the present invention, in minimizing interference between containers, the method includes:

[0016] Calculate the interference index of each container and classify the containers according to the interference type; the container interference type includes computing resource interference, network resource interference and I / O resource interference;

[0017] For each container, select a candidate node with the least interference;

[0018] For high-interference containers, a target optimization algorithm is used to optimize to minimize the interference between containers.

[0019] In some embodiments of the present invention, the method further comprises:

[0020] Monitor the containers running on each node in real time to obtain the current interference index changes;

[0021] When the interference index on any node exceeds the set threshold, scheduling is retriggered to migrate high-interference containers to other nodes.

[0022] In some embodiments of the present invention, the method further comprises:

[0023] By analyzing historical scheduling data through machine learning algorithms, the best scheduling solution for minimizing interference between containers in different scenarios is found and integrated into the scheduler to implement an optimized feedback mechanism.

[0024] A microservice scheduling system based on cloud-edge collaboration is proposed, including:

[0025] The task decomposition unit is used to obtain user task requests and decompose them into multiple microservices to be scheduled;

[0026] A node screening unit is used to screen edge nodes and / or cloud nodes that meet the resource requirements of each microservice to be scheduled, and build a candidate node set;

[0027] The target model building unit is used to build the target function with the goal of minimizing service delay time, minimizing scheduling time and minimizing interference between containers; wherein the model for minimizing interference between containers is: , To calculate the interference index, is the CPU resource requested by the container, is the CPU resources available to the node; is the network interference index, is the bandwidth consumed by the container, It is the I / O resources available to the node; = is the I / O interference index, are the I / O resources requested by the container, is the I / O resources available to the node;

[0028] The optimal microservice scheduling calculation unit is used to solve the objective function and obtain the optimal microservice scheduling result.

[0029] In some embodiments of the present invention, the node screening unit is further used to:

[0030] Generate a microservice task graph, where nodes represent microservices and edges represent dependencies between microservices;

[0031] For each microservice to be scheduled, minimize cross-cluster or cross-host calls between microservices.

[0032] In some embodiments of the present invention, in minimizing interference between containers, the optimal microservice scheduling calculation unit is used to:

[0033] Calculate the interference index of each container and classify the containers according to the interference type; the container interference type includes computing resource interference, network resource interference and I / O resource interference;

[0034] For each container, select a candidate node with the least interference;

[0035] For high-interference containers, a target optimization algorithm is used to optimize to minimize the interference between containers.

[0036] In some embodiments of the present invention, the system further comprises:

[0037] The adaptive scheduling unit is used to monitor the containers running on each node in real time and obtain the current interference index changes; when the interference index on any node exceeds the set threshold, the scheduling is re-triggered to migrate the high-interference containers to other nodes.

[0038] In some embodiments of the present invention, the system further comprises:

[0039] The interference learning and optimization feedback unit is used to analyze historical scheduling data through machine learning algorithms, find the best scheduling solution for minimizing interference between containers in different scenarios, and integrate it into the scheduler to implement an optimized feedback mechanism.

[0040] Compared with the prior art, the advantages and positive effects of the present invention are as follows: in the microservice scheduling method and system based on cloud-edge collaboration proposed in the present invention, for the intelligent industrial process scenario, the user task request is decomposed into multiple microservices to be scheduled, and for each microservice to be scheduled, the edge nodes and / or cloud nodes that meet its resource requirements are screened to construct a candidate node set, the interference types between containers are subdivided and quantified, and an interference index calculation model is constructed for each type of interference between containers. The objective function is constructed with the goals of minimizing interference between containers, improving cluster resource utilization, and reducing service response time, thereby converting the microservice scheduling problem into a multi-objective optimization problem, which not only realizes the reasonable scheduling of microservices in the intelligent monitoring scenario of industrial processes, but also reduces interference between containers and improves the service quality of edge applications.

[0041] Other features and advantages of the present invention will become more apparent after reading the detailed description of the embodiments of the present invention in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0043] Figure 1 The following is a schematic diagram of the execution steps of the microservice scheduling method based on cloud-edge collaboration proposed by the present invention;

[0044] Figure 2 This is a schematic diagram of the cloud-edge collaborative industrial process intelligent monitoring microservice system mechanism in the present invention;

[0045] Figure 3 This is an example of microservice scheduling for intelligent industrial processes in the present invention;

[0046] Figure 4 This is a schematic diagram of the system structure of the microservice scheduling system based on cloud-edge collaboration proposed in the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] In the description of the present invention, it should be understood that the terms "center", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship are based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0049] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0050] The terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "plurality" means two or more.

[0051] Industrial process operation monitoring is the basis for ensuring efficient, safe and reliable operation of production systems. However, the manufacturing process mechanism is complex and unclear, key indicators are coupled and correlated, and intelligent autonomous operation monitoring modeling is difficult under multi-constraint global optimization conditions. Existing monitoring and data acquisition systems often rely on variable out-of-limit or manual experience, and there are problems such as inaccurate judgment of operation monitoring anomalies and untimely decision-making. In addition, strongly coupled complex models are difficult to deploy and apply efficiently under low edge computing power and high reliability requirements, and lack a cloud-edge collaboration mechanism.

[0052] In response to the needs of business collaboration and component service-oriented in complex cloud-edge network environments, and improving component model combination configuration and model sharing capabilities driven by business needs, the present invention aims to study the service-oriented construction technology of component models for industrial process operation monitoring, and the model assembly and integration mode for business problem subdomains. By analyzing the characteristics of industrial process operation monitoring business scenarios, an event-driven component model and service mapping relationship is established, and the algorithm component model provides services to the outside through a three-layer service encapsulation system of model definition layer, model service layer and model publishing layer, realizing multi-granularity decoupling and service encapsulation of algorithm components, establishing a cloud-edge collaboration mechanism, and constructing, publishing, iterating and managing cloud-edge collaborative microservices. Develop microservice construction tools for efficient application of intelligent algorithms, build cloud-edge collaborative microservice programs for intelligent operation monitoring in typical scenarios, form a resource pool of intelligent operation monitoring microservice programs under distributed clusters, realize cloud-edge microservice collaborative application in typical scenarios of industrial process operation monitoring, and open up the collaborative work of cloud-edge resources, data, intelligent models, applications, business management, services and other elements to solve the current problems in industrial scenarios. The existing heterogeneous resources (such as gateways, cameras, etc.) are numerous, scattered and not unified, which is not conducive to management by managers. In addition, the traditional resource construction method based on virtual machines or bare metal has high overhead, and existing resources are not fully utilized. The current microservice scheduling research mainly exists in the cloud environment, and does not take into account the interference of differences between edge computing scenarios and cloud computing network environments, making migration difficult.

[0053] In combination with the above, the present invention designs an edge computing resource virtualization framework for smart parks, builds an edge resource pool, uses lightweight container technology, and combines with an open source edge computing framework to construct the edge resources of the park, develops an embedded edge computing gateway with edge computing and virtualization capabilities, and designs a smart park edge service system to manage and schedule resources to improve overall resource utilization.

[0054] A cloud-edge collaborative microservice scheduling algorithm for smart park scenarios is proposed. A smart park edge service system is designed based on the smart park scenario. According to the characteristics of the park scenario, the microservice problem is transformed into a multi-objective optimization problem with the goal of improving the overall resource utilization of the park, reducing interference between containers, and reducing the cross-host call distance of microservice containers. The problem is solved using the proposed microservice scheduling algorithm, and the algorithm is integrated into the smart park edge service system.

[0055] Optimize the intelligent algorithms in the edge computing environment and design lightweight algorithm models that are suitable for low computing resource environments to ensure that complex intelligent algorithms can run efficiently on edge devices.

[0056] By introducing quantitative analysis of interference types, interference perception strategies, real-time monitoring and dynamic adjustment, and an optimization feedback mechanism based on machine learning, we can further refine the interference minimization problem, avoid the general interference handling methods in existing technologies, improve the accuracy of scheduling, adapt to complex resource demands and service quality requirements in different industrial scenarios, and significantly improve the overall performance and reliability of the system.

[0057] The present invention describes the proposed microservice scheduling method and system based on cloud-edge collaboration from the following aspects.

[0058] 1. Description of cloud-edge collaborative microservice scheduling.

[0059] In the intelligent industrial process scenario, when a user initiates a task request, multiple microservices need to work together to meet the user's task requirements. A reasonable microservice scheduling algorithm is needed to schedule the container carrying the microservice to the physical node in the edge cloud or central cloud for deployment to provide services to users and improve service quality.

[0060] Some current open source container orchestration platforms only provide some simple scheduling strategies, without considering the network conditions in the industrial process and the dependencies between microservice containers. Unreasonable scheduling strategies will lead to waste of cluster resources and degraded service quality.

[0061] Some edge applications in industrial processes (such as face recognition, electronic inspection, license plate recognition, etc.) have strict requirements on network bandwidth and time, and require rapid deployment. Traditional cloud computing methods can no longer meet the needs. Therefore, this paper uses a cloud-edge collaborative solution based on a container-based microservice application architecture to study the scheduling of microservices in intelligent industrial processes. Figure 1 shown.

[0062] The present invention receives task requests initiated by users through a custom scheduler deployed on the Master node in the central cloud, combines the node status of the edge cloud and the central cloud, and uses the proposed microservice scheduling algorithm based on cloud-edge collaboration to reasonably schedule the microservices to be scheduled. The microservices that are allowed to be scheduled to the edge side are deployed in the corresponding edge nodes, making full use of edge side resources, providing services nearby, and improving service quality.

[0063] There are mutual calling relationships between microservices. Therefore, in the actual scheduling process, cross-host or cross-cluster microservice calls may occur. Cross-cluster calls between microservices should be minimized. An example of smart industrial process microservice scheduling is as follows: Figure 2 shown.

[0064] Containers are the carriers of microservices. Microservices correspond to containers one by one. A Pod can deploy one or more containers. Cloud-edge collaborative microservice scheduling is to deploy the Pod to be scheduled to the appropriate physical node. The user sends the task request in the form of a JSON file or a YAML file to the scheduler deployed in the central cloud. The scheduler uses the proposed scheduling algorithm to convert the microservice scheduling problem into a multi-objective optimization problem, solve it, and calculate the scheduling result to avoid deploying the Pod to nodes where resources such as bandwidth, memory, and CPU are saturated. The scheduling results will be stored in the database. The physical nodes in the edge cloud or central cloud query the scheduling results in the database, create corresponding containers, and run the corresponding microservices to provide services to users.

[0065] 2. Analysis of scheduling factors for cloud-edge collaborative microservices.

[0066] The microservice scheduling based on cloud-edge collaboration proposed in the present invention means: according to the task request initiated by the user, the scheduler deployed in the central cloud schedules the Pod carrying the microservice container according to the constraints, and creates the corresponding container on the physical node of the edge cloud or the central cloud according to the scheduling result. The key elements of the cloud-edge collaborative microservice scheduling problem are analyzed below.

[0067] Constraints: There are hard constraints and soft constraints. Hard constraints refer to constraints that are not allowed to be violated during the task scheduling process, mainly including the order of tasks, execution time, task processor constraints, upper and lower limits of resources required during task execution, upper and lower limits of hardware resources during scheduling, etc. Soft constraints refer to constraints that are met as much as possible during the task scheduling process, mainly including user preferences, execution time constraints, task processor constraints, etc.

[0068] Objective function: The objective function of cloud-edge collaborative microservice scheduling is a multi-objective optimization function. Its purpose is to minimize scheduling time, service delay time, and interference between containers through a reasonable scheduling strategy. The objective function is set as follows:

[0069] Goal 1, minimize service delay time. In the application scenario of intelligent industrial process, some edge applications have strict requirements on network delay, such as face recognition, electronic inspection, etc., which require rapid deployment, and traditional cloud computing methods can no longer meet the needs. In the design of this invention, some microservices are deployed at the edge of the network, and users are provided with nearby services through edge computing to improve service quality. Service delay time refers to the time required from the user initiating a request to the microservice returning the result. The service delay time is defined as follows:

[0070]

[0071] in, Indicates the service delay time between microservices; and Represents a microservice; Representing microservices With Microservices The service delay time between .

[0072] Goal 2, minimize scheduling time. Microservice scheduling refers to the reasonable scheduling of task requests initiated by users to physical nodes of edge cloud or central cloud. Reasonable scheduling can not only improve cluster resource utilization, but also reduce scheduling time and improve system response speed. Scheduling time refers to the time required from the user initiating a task request to the storage of the scheduling result in the database. The definition of scheduling time is as follows:

[0073]

[0074] in, Indicates the service delay time between microservices; and Represents a microservice; Representing microservices With Microservices The scheduling time between.

[0075] Objective 3, minimize the interference between containers. Minimizing interference is a key optimization goal in microservice scheduling, but common research and methods do not refine the specific types of interference and quantitative standards. In this invention, the following innovative means are implemented to minimize interference:

[0076] (1) Subdivide and quantify interference types.

[0077] Traditional microservice scheduling solutions focus on the scheduling of basic resources such as CPU and memory, but do not systematically analyze and quantify the interference between containers. In this paper, the interference between containers is subdivided into the following categories:

[0078] 1. Computing resource interference: Multiple containers compete for CPU, GPU and other hardware resources at the same time, causing some containers to be unable to obtain resources in time, affecting their performance.

[0079] 2. Network resource interference: Multiple containers exchange large amounts of data at the same time, resulting in network bandwidth being occupied, which in turn affects the response speed of the service.

[0080] 3. I / O resource interference: When containers frequently read and write disks or other peripherals, operation conflicts among multiple containers may cause I / O operations to be blocked.

[0081] The present invention defines corresponding metrics for each type of interference:

[0082] 1. Compute Interference Index (CII): It indicates the degree of competition for CPU resources among different containers on the same node. The calculation formula is as follows:

[0083] ;

[0084] in, is the CPU resource requested by the container, is the CPU resources available to the node.

[0085] 2. Network Interference Index (NII): indicates the network traffic interference between containers. The calculation formula is as follows:

[0086] ;

[0087] in, is the bandwidth consumed by the container, It is the I / O resources available to the node.

[0088] 3. I / O Interference Index (IOII): Indicates the contention of containers for I / O resources. The calculation formula is as follows:

[0089] = ;

[0090] are the I / O resources requested by the container, Is the I / O resources available to the node.

[0091] Through these quantitative indicators, the objective function of interference minimization can be defined as:

[0092] ;

[0093] This objective function provides a more precise optimization direction by integrating the interference conditions of multiple types of resources.

[0094] (2) Plan interference-aware scheduling strategies.

[0095] In order to further reduce the interference between containers, the present invention designs a scheduling algorithm based on interference perception. According to the above interference index, when scheduling microservice containers, the algorithm tries to allocate high-interference containers to different physical nodes to avoid mutual influence. The steps are as follows:

[0096] 1. Calculate the interference index of each container and classify the containers according to the type of interference.

[0097] 2. Select a candidate node for each container, giving priority to the node with the least interference.

[0098] 3. For high-interference containers, use a multi-objective optimization algorithm (such as genetic algorithm or particle swarm optimization) for further optimization to ensure that the containers are reasonably distributed and the interference is minimized.

[0099] (3) Dynamic interference monitoring and real-time scheduling.

[0100] A dynamic interference monitoring system is introduced to monitor the containers running on each node in real time to obtain the current interference index changes in a timely manner. Once the interference index on a node exceeds the set threshold, the system will trigger a rescheduling mechanism to migrate high-interference containers to other nodes. The features of this dynamic scheduling system include:

[0101] Real-time monitoring: Continuously monitor the usage of CPU, network, I / O and other resources to ensure that the system can respond in a timely manner.

[0102] Adaptive scheduling: Combine historical data with current monitoring information to predict interference trends and adjust container distribution in advance to prevent interference surges.

[0103] (4) Interference learning and optimization feedback mechanism.

[0104] Through machine learning algorithms, historical scheduling data is analyzed to find the best interference minimization strategies for different scenarios. These strategies are integrated into the scheduler to form a continuously optimized feedback mechanism, ensuring that the scheduling system can gradually improve the rationality and efficiency of container allocation after long-term operation.

[0105] 3. Cloud-edge collaborative microservice scheduling algorithm.

[0106] In order to achieve efficient microservice scheduling in the intelligent industrial process, this paper designs a multi-objective optimization scheduling algorithm based on cloud-edge collaboration. The algorithm combines the advantages of edge computing and cloud computing, and achieves efficient deployment and operation of microservices through reasonable scheduling strategies.

[0107] Combination Figure 1 As shown, the microservice scheduling algorithm based on cloud-edge collaboration of the present invention includes:

[0108] 1. Initialization, including:

[0109] Input: includes user task requests, resource information of edge nodes and cloud nodes, and microservice dependencies.

[0110] Output: Microservice scheduling results.

[0111] Scheduler: Get information about all currently available edge nodes and cloud nodes.

[0112] 2. Task decomposition:

[0113] (1) Decompose user task requests into multiple microservices, determine the resource requirements (CPU, memory, bandwidth, etc.) of each microservice and the dependencies between microservices.

[0114] (2) Generate a microservice task graph, where the nodes represent microservices and the edges represent the dependencies between microservices.

[0115] 3. Node screening:

[0116] (1) For each microservice to be scheduled, select edge nodes and cloud nodes that meet the resource requirements and build a set of candidate nodes.

[0117] (2) For microservices with high real-time requirements, edge nodes are preferred; for microservices with large computing resource requirements, cloud nodes are preferred.

[0118] 4. Construction of multi-objective optimization model.

[0119] The objective functions include: (1) minimizing service delay time :

[0120] ;

[0121] (2) Minimize scheduling time :

[0122] ;

[0123] (3) Minimize interference between containers :

[0124] .

[0125] Constraints include: (1) Hard constraints: resource limitations, microservice dependencies, task priorities, etc.

[0126] (2) Soft constraints: service quality requirements, user preferences, etc.

[0127] 5. Scheduling algorithm design.

[0128] (1) Ant Colony Optimization (ACO):

[0129] (a) Initialize the ant colony, where each ant represents a possible scheduling solution.

[0130] (b) According to the microservice task graph and resource information, the ants start to facilitate the task graph from the starting node and select the appropriate node for microservice scheduling.

[0131] (c) Through the pheromone updating mechanism, the pheromone concentration of high-quality scheduling plans is increased, while the pheromone concentration of low-quality plans is reduced.

[0132] (2) Genetic Algorithm (GA):

[0133] (a) Initialize the population, where each individual represents a possible scheduling solution.

[0134] (b) The scheduling scheme is continuously optimized through selection, crossover, and mutation operations until the termination condition is met (such as reaching the maximum number of iterations or meeting the service quality requirements).

[0135] (3) Particle Swarm Optimization (PSO):

[0136] (a) Initialize the particle swarm, where each particle represents a possible scheduling solution.

[0137] (b) The particle moves in the search space, updates its speed and position according to its own and the global optimal position, and gradually approaches the optimal solution.

[0138] 6. Scheduling results are generated.

[0139] (1) Generate the final microservice scheduling plan based on the output results of the optimization algorithm.

[0140] (2) The scheduling results are stored in the database for query by edge nodes and cloud nodes.

[0141] As mentioned above, the present invention designs an edge computing resource virtualization architecture for intelligent industrial processes, combines lightweight container technology and open source edge computing framework, builds an efficient edge service system, and uniformly manages and schedules edge resources. The proposed microservice scheduling algorithm based on cloud-edge collaboration improves resource utilization in edge computing environments through multi-objective optimization strategies, reduces interference between containers and service delay time, and improves the service quality of intelligent monitoring of industrial processes. It has important application value in improving the flexibility and response speed of intelligent industrial processes.

[0142] In combination with the above-mentioned microservice scheduling method based on cloud-edge collaboration, the present invention also proposes a microservice scheduling system based on cloud-edge collaboration, such as Figure 4 As shown, including:

[0143] The task decomposition unit 1 is used to obtain the user task request and decompose it into multiple microservices to be scheduled.

[0144] The node screening unit 2 is used to screen edge nodes and / or cloud nodes that meet the resource requirements of each microservice to be scheduled, and build a candidate node set.

[0145] The target model building unit 3 is used to build an objective function with the goal of minimizing service delay time, minimizing scheduling time and minimizing interference between containers; wherein the model for minimizing interference between containers is: , To calculate the interference index, is the CPU resource requested by the container, is the CPU resources available to the node; is the network interference index, is the bandwidth consumed by the container, It is the I / O resources available to the node; = is the I / O interference index, are the I / O resources requested by the container, Is the I / O resources available to the node.

[0146] The optimal microservice scheduling calculation unit 4 is used to solve the objective function to obtain the optimal microservice scheduling result.

[0147] The adaptive scheduling unit 5 is used to monitor the containers running on each node in real time and obtain the current interference index change; when the interference index on any node exceeds the set threshold, the scheduling is re-triggered to migrate the high-interference container to other nodes.

[0148] The interference learning and optimization feedback unit 6 is used to analyze the historical scheduling data through a machine learning algorithm, find the best scheduling scheme for minimizing the interference between containers in different scenarios, and integrate it into the scheduler to realize the optimized feedback mechanism.

[0149] As mentioned above, the node screening unit 2 is also used to: generate a microservice task graph, in which the nodes represent microservices and the edges represent the dependencies between microservices; and for each microservice to be scheduled, minimize cross-cluster or cross-host calls between microservices.

[0150] In some embodiments of the present invention, in minimizing interference between containers, the optimal microservice scheduling calculation unit 4 is used to: calculate the interference index of each container, and classify the containers according to the type of interference; wherein the container interference types include computing resource interference, network resource interference, and I / O resource interference; select a candidate node with the least interference for each container; for high-interference containers, use a target optimization algorithm to optimize to minimize interference between containers.

[0151] In some embodiments of the present invention, in minimizing service delay time, the optimal microservice scheduling calculation unit 4 is used to: deploy microservices with high network latency requirements at the edge of the network, and provide users with nearby services through edge computing.

[0152] It should be noted that, in the specific implementation process, the above-mentioned control part can be implemented by a hardware processor executing computer execution instructions in software form stored in the memory, which will not be elaborated here, and the programs corresponding to the actions performed by the above-mentioned control circuit can be stored in the computer-readable storage medium of the system in software form, so that the processor can call and execute the operations corresponding to the above-mentioned modules.

[0153] The computer-readable storage medium mentioned above may include volatile memory, such as random access memory; may also include non-volatile memory, such as read-only memory, flash memory, hard disk or solid-state drive; may also include a combination of the above types of memory.

[0154] The processor mentioned above may also be a general term for multiple processing elements. For example, the processor may be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits, field programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc., or may be a dedicated processor.

[0155] It should be pointed out that the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by ordinary technicians in this technical field within the essential scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A microservice scheduling method based on cloud-edge collaboration, characterized in that: include: Get user task requests and decompose them into multiple microservices to be scheduled; For each microservice to be scheduled, select edge nodes and / or cloud nodes that meet its resource requirements and build a candidate node set; The objective function is constructed with the goal of minimizing service delay time, minimizing scheduling time, and minimizing interference between containers. The model for minimizing interference between containers is: , To calculate the interference index, is the CPU resource requested by the container, is the CPU resources available to the node; is the network interference index, is the bandwidth consumed by the container, It is the I / O resources available to the node; = is the I / O interference index, are the I / O resources requested by the container, is the I / O resources available to the node; Solve the objective function to obtain the optimal microservice scheduling result.

2. The microservice scheduling method based on cloud-edge collaboration according to claim 1 is characterized in that: After decomposing to obtain a plurality of microservices to be scheduled, the method further includes: Generate a microservice task graph, where nodes represent microservices and edges represent dependencies between microservices; For each microservice to be scheduled, minimize cross-cluster or cross-host calls between microservices.

3. The microservice scheduling method based on the cloud-edge system according to claim 1 is characterized in that: In minimizing interference between containers, the method includes: Calculate the interference index of each container and classify the containers according to the interference type; the container interference type includes computing resource interference, network resource interference and I / O resource interference; For each container, select a candidate node with the least interference; For high-interference containers, a target optimization algorithm is used to optimize to minimize the interference between containers.

4. The microservice scheduling method based on cloud-edge collaboration according to claim 1 is characterized in that: The method further comprises: Monitor the containers running on each node in real time to obtain the current interference index changes; When the interference index on any node exceeds the set threshold, scheduling is retriggered to migrate high-interference containers to other nodes.

5. The microservice scheduling method based on cloud-edge collaboration according to claim 1 is characterized in that: The method further comprises: By analyzing historical scheduling data through machine learning algorithms, the best scheduling solution for minimizing interference between containers in different scenarios is found and integrated into the scheduler to implement an optimized feedback mechanism.

6. A microservice scheduling system based on cloud-edge collaboration, characterized by: include: The task decomposition unit is used to obtain user task requests and decompose them into multiple microservices to be scheduled; A node screening unit is used to screen edge nodes and / or cloud nodes that meet the resource requirements of each microservice to be scheduled, and build a candidate node set; The target model building unit is used to build the target function with the goal of minimizing service delay time, minimizing scheduling time and minimizing interference between containers; wherein the model for minimizing interference between containers is: , To calculate the interference index, is the CPU resource requested by the container, is the CPU resources available to the node; is the network interference index, is the bandwidth consumed by the container, It is the I / O resources available to the node; = is the I / O interference index, are the I / O resources requested by the container, is the I / O resources available to the node; The optimal microservice scheduling calculation unit is used to solve the objective function and obtain the optimal microservice scheduling result.

7. The microservice scheduling system based on cloud-edge collaboration according to claim 6 is characterized in that: The node screening unit is also used for: Generate a microservice task graph, where nodes represent microservices and edges represent dependencies between microservices; For each microservice to be scheduled, minimize cross-cluster or cross-host calls between microservices.

8. The microservice scheduling system based on cloud-edge collaboration according to claim 6 is characterized in that: In minimizing the interference between containers, the optimal microservice scheduling computing unit is used to: Calculate the interference index of each container and classify the containers according to the interference type; the container interference type includes computing resource interference, network resource interference and I / O resource interference; For each container, select a candidate node with the least interference; For high-interference containers, a target optimization algorithm is used to optimize to minimize the interference between containers.

9. The microservice scheduling system based on cloud-edge collaboration according to claim 6 is characterized in that: The system further comprises: The adaptive scheduling unit is used to monitor the containers running on each node in real time and obtain the current interference index changes; when the interference index on any node exceeds the set threshold, the scheduling is re-triggered to migrate the high-interference containers to other nodes.

10. The microservice scheduling system based on cloud-edge collaboration according to claim 6, characterized in that: The system further comprises: The interference learning and optimization feedback unit is used to analyze historical scheduling data through machine learning algorithms, find the best scheduling solution for minimizing interference between containers in different scenarios, and integrate it into the scheduler to implement an optimized feedback mechanism.

Images