Task-level cooperative optimization dispatching method supporting multi-cloud network disconnection disaster recovery

Abstract

The present invention relates to a task-level cooperative optimization dispatching method supporting multi-cloud network disconnection disaster recovery. On the basis of the prior art, for the problem that cooperative dispatching is required to be performed according to different resource requirements for tasks in a network anomaly stage, nodes in a production workshop form a cooperative computing system, production tasks are transmitted and processed among the nodes according to process requirements, and task-level cooperative optimization dispatching under a network disconnection condition is performed by means of a CoopDispatch method, ensuring continuous production. The CoopDispatch method is used, for the high specific resource-dependency characteristic of manufacturing tasks, transmission, storage and computing of data are implemented by means of a LocalAny mechanism, reducing the time delay of the tasks, improving the cooperative efficiency between the tasks, and satisfying the requirements of manufacturing enterprises for network disconnection disaster recovery in a multi-cloud mode.

Description

A task-level collaborative optimization scheduling method supporting multi-cloud network outage disaster recovery Technical Field

[0001] The present invention relates to the field of computer system information technology, and in particular to a task-level collaborative optimization scheduling method that supports multi-cloud network outage disaster recovery. Background Art

[0002] The rapid development of next-generation information technology provides strong technical support for technological integration and innovation in the manufacturing industry. The key development directions of the new round of industrial revolution are to promote high-quality development of the manufacturing industry, strengthen the industrial foundation and technological innovation capabilities, promote the integrated development of advanced manufacturing and modern service industries, accelerate the development of a manufacturing powerhouse, build industrial Internet platforms, and expand "intelligence plus" to empower the transformation and upgrading of the manufacturing industry. Against this backdrop, intelligent enterprise decision-making requires new application innovation platforms. With the convergence of manufacturing transformation and the digital economy, the integrated innovation of information technologies such as cloud computing, the Internet of Things, and big data, along with manufacturing technologies and industrial knowledge, is intensifying, giving rise to industrial Internet platforms. Migrating enterprise services to the cloud has become a trend.

[0003] In the traditional manufacturing model, the company's business system runs locally and is not affected by peripheral network failures. However, after the company's business system is put on the cloud, if it continues to use the previous "resource-centric" model of client request and server response, once the network between the enterprise and the cloud fails, it will affect the company's business system. Based on the existing technology, the present invention focuses on solving the problem of task collaborative optimization caused by the high dependence on specific resources of manufacturing tasks when the local area network of the production workshop is normal but the network between multiple clouds is disconnected in the multi-cloud model, filling the technical gap in this field.

[0004] Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a task-level collaborative optimization scheduling method that supports multi-cloud network disconnection disaster recovery. Taking into account that the "resource-centric" real-time request-response mode of traditional production control systems is not applicable to multi-cloud environments, in the production process control system under multi-cloud environments, a large amount of frequent data interaction is required between manufacturing nodes. The method of the present invention provides task-level collaborative optimization scheduling through the CoopDispatch method in the event of a network disconnection to ensure continuous production; the LocalAny mechanism is used to realize data transmission, storage, and calculation, which can reduce task latency, improve the collaborative efficiency between tasks, and meet the needs of manufacturing enterprises for network disconnection disaster recovery in a multi-cloud mode.

[0006] The technical solution adopted by the present invention to achieve the above-mentioned purpose is: a task-level collaborative optimization scheduling method supporting multi-cloud network outage disaster recovery, comprising the following steps:

[0007] Manufacturing nodes assign tasks to other matching manufacturing nodes through a non-service-dependent end-to-end communication channel;

[0008] The CoopDispatch method is used between manufacturing nodes to coordinate and optimize the scheduling of tasks with other manufacturing nodes;

[0009] The task processing results of a manufacturing node and other manufacturing nodes are synchronized between the manufacturing nodes through the LocalAny mechanism.

[0010] The manufacturing node is a client used by the edge side of a multi-cloud environment to host a production control system in a manufacturing enterprise workshop.

[0011] The multi-cloud includes the cloud formed after the enterprise management and control business is moved to the cloud, and the sub-clouds of the production workshops of various manufacturing enterprises.

[0012] The CoopDispatch method calculates the minimum average execution time of resource-constrained tasks in a collaborative computing system composed of multiple manufacturing nodes, and collaboratively schedules tasks to improve the overall benefits of the collaborative computing system. The method includes the following steps:

[0013] 1) The tasks in the collaborative computing system CS are divided into |M| categories, and the task set is represented as Task. For the mth category task, m∈|M|, its task is only in the resource R i Task collection m ∈Task, for the current manufacturing node, the arrival rate of the mth type task request to other manufacturing nodes at time t satisfy:

[0014] in, represents the maximum arrival rate of the mth task, i represents the i-th task of the mth task, Is a flag used to indicate resource R i Tasks that can be processed T i Is it in the task collection Task m In Chinese, that is:

[0015] The manufacturing node processes multiple tasks simultaneously, and the mth task at time t is stored in In the queue, all task queues The following queue matrix Q(t) is constructed:

[0016] r i Represents a resource collection R i A resource, It represents the tail task of the queue with the number of waiting times |M| for processing the m-th type of task; Represents resource r i The mth type of task processed, i=1…I;

[0017] The number of task requests of type m received at time t is defined as It satisfies:

[0018] Among them, Δt represents the time interval and is a constant;

[0019] The average task execution time is as follows:

[0020] Among them, α is the empirical constant, E{} represents the expectation, and here represents the average time of task execution expect;

[0021] 2) The collaborative computing system CS allocates resources required to process tasks. The resource CS corresponding to the i-th task i The resources allocated to the mth task request are

[0022] The manufacturing node updates the task processing queue through the following model:

[0023] in, Indicates the number of tasks of type m that a unit resource can process within a set period.

[0024] 3) Based on the average task execution time and resources Construct a comprehensive benefit model to maximize the comprehensive benefits of the collaborative computing system CS to obtain the minimum average execution time and resources to achieve collaborative optimization scheduling.

[0025] In step 3), the collaborative optimization scheduling is achieved by maximizing the comprehensive benefits of the collaborative computing system CS, including the following steps:

[0026] (1) When a manufacturing node's task is accepted by other manufacturing nodes, the manufacturing node obtains the following benefits:

[0027] in, represents the first part of the benefit, β m is the m-th task request benefit constant, is the average time for task execution;

[0028] (2) The number of task requests of type m processed by the manufacturing node at time t, n m (t) is:

[0029] Indicates the number of tasks of type m that can be processed by a unit resource within a set period;

[0030] The corresponding benefits are:

[0031] Among them, n m n m (t) time expectation, θ m is the corresponding profit constant; represents the second part of the benefit, which depends on the number of requests served;

[0032] (3) The comprehensive benefit model is:

[0033] Among them, i represents the i-th task of the m-th category;

[0034] (4) In order to maximize the comprehensive benefits and achieve collaborative optimization, when B reaches its maximum value, the obtained is the minimum average execution time.

[0035] The task processing results of the manufacturing node and other manufacturing nodes are synchronized between the manufacturing nodes through the LocalAny mechanism, including the following steps:

[0036] When a manufacturing node receives tasks from other nodes, the first part of the benefit It is nonlinear, adding a transformation constant variable and constant variable transpose queue Convert task optimization scheduling into a linear problem:

[0037] Among them, w m and is a constant variable, Φ(Ω(t)) represents the maximum comprehensive benefit value, Ω(t) represents the difference between the current comprehensive benefit and the average comprehensive benefit, E{} represents the expectation, and here represents the expected value of the benefit. is the average time for task execution;

[0038] Maximize the average execution time obtained by the comprehensive benefit of the collaborative computing system CS in step 3) and resources Substitute the combination into Φ(Ω(t)) and obtain the optimized task scheduling strategy by minimizing Φ(Ω(t)):

[0039] When the real queue When the queue is not longer than the constant variable transposition queue, the manufacturing node accepts all requests for the mth type of tasks arriving at time t, otherwise it waits.

[0040] The present invention has the following beneficial effects and advantages:

[0041] 1. The method of the present invention uses the CoopDispatch method to perform task-level collaborative optimization scheduling in the event of a network outage, which can ensure continuous production;

[0042] 2. The LocalAny mechanism adopted by the present invention realizes data transmission, storage, and calculation, which can reduce task latency, improve the collaborative efficiency between tasks, and meet the needs of manufacturing enterprises for network outage disaster recovery in a multi-cloud mode. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] FIG1 is a schematic diagram of the method of the present invention. DETAILED DESCRIPTION

[0044] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0045] FIG1 is a schematic diagram of the method of the present invention.

[0046] The present invention relates to a task-level collaborative optimization scheduling method that supports multi-cloud network disconnection disaster recovery. With the continuous development and popularization of cloud computing technology, it has become a trend for manufacturing enterprises to go to the cloud. After the enterprise goes to the cloud, when the local area network of the production workshop is normal but the network between multiple clouds is disconnected, how to ensure normal and continuous production has become a technical issue of concern to the industry; in view of the different resource requirements for tasks in the network abnormality stage, collaborative scheduling is required. The nodes in the production workshop form a collaborative computing system, and the production tasks are transmitted and processed between the nodes according to the process requirements. The CoopDispatch method is used to perform task-level collaborative optimization scheduling in the case of network disconnection to ensure continuous production; the CoopDispatch method adopted in the present invention is based on the high dependence on specific resources of manufacturing tasks. The data transmission, storage and calculation are realized through the LocalAny mechanism, which can reduce the task latency, improve the collaborative efficiency between tasks, and meet the needs of manufacturing enterprises for network disconnection disaster recovery in a multi-cloud mode.

[0047] A task-level collaborative optimization scheduling method supporting multi-cloud network outage disaster recovery includes the following steps:

[0048] Manufacturing nodes allocate computationally intensive and resource-sensitive tasks to matching manufacturing nodes through a non-service-dependent end-to-end communication channel;

[0049] Computational intensive refers to the high computing resources required to process the task, such as GPU resources, high computing power and other special resources;

[0050] Resource-sensitive refers to tasks that require specific manufacturing resources, such as equipment with special production capacity or production lines with specific processes based on the production process;

[0051] The CoopDispatch method is used to coordinate and optimize the scheduling of tasks between manufacturing nodes.

[0052] The processing results of the task are synchronized among the manufacturing nodes through the LocalAny mechanism.

[0053] The manufacturing node is a client used by the edge side of a multi-cloud environment to host a production control system in a manufacturing enterprise workshop.

[0054] The multi-cloud includes the cloud formed after the enterprise management and control business is moved to the cloud, and the sub-clouds of the production workshops of various manufacturing enterprises.

[0055] The CoopDispatch method calculates the minimum average execution time of resource-constrained tasks in a collaborative computing system and coordinates the scheduling of tasks to improve the overall benefits of the system. The method includes the following steps:

[0056] The tasks in the collaborative computing system CS are divided into |M| categories, and the task set is represented as Task. For the mth category task, m∈|M|, its task is only in the resource R i Task collection m ∈Task, for the current manufacturing node, the arrival rate of the mth type task request to other manufacturing nodes at time t satisfy:

[0057] in, represents the maximum arrival rate of the mth task, i represents the i-th task of the mth task, Is a flag used to indicate resource R i Tasks that can be processed T i Is it in the task collection Task m In Chinese, that is:

[0058] The manufacturing node processes multiple tasks simultaneously, and the mth task at time t is stored in In the queue, all task queues The following queue matrix Q(t) is constructed:

[0059] r i Represents a resource collection R i A resource, It represents the tail task of the queue with the number of waiting times |M| for processing the m-th type of task; Represents resource ri The mth type of task queue to be processed, i=1…I;

[0060] The number of task requests of type m received at time t is defined as It satisfies:

[0061] Among them, Δt represents the time interval and is a constant;

[0062] The average task execution time is as follows:

[0063] Among them, α is the empirical constant, E represents the average time of task execution expect;

[0064] The collaborative computing system CS allocates the resources required to process the tasks. The resource CS corresponding to the i-th task is i The resources allocated to the mth task request are

[0065] The manufacturing node updates the task processing queue through the following model:

[0066] in, Indicates the number of tasks of type m that can be processed by a unit resource within a certain period.

[0067] Based on the average task execution time and resources Construct a comprehensive benefit model to maximize the comprehensive benefits of the collaborative computing system CS to obtain the minimum average execution time and resources to achieve collaborative optimization scheduling.

[0068] Finally, the goal of collaborative optimization is achieved by maximizing the comprehensive benefits of the collaborative computing system CS. The benefits can be divided into two parts:

[0069] When a manufacturing node's tasks are accepted by other manufacturing nodes, the manufacturing node obtains the following benefits:

[0070] in, represents the first part of the benefit, β m is the m-th task request benefit constant, is the average time for task execution;

[0071] The number of task requests of type m processed by the manufacturing node at time t, n m (t) is:

[0072] The corresponding benefits are:

[0073] Among them, n m n m (t) time expectation, θ m is the corresponding profit constant; represents the second part of the benefit, which depends on the number of requests served;

[0074] The comprehensive benefit model is:

[0075] Among them, i represents the i-th task of the m-th category;

[0076] In order to maximize the comprehensive benefits and achieve collaborative optimization, when B reaches its maximum value, the obtained is the minimum average execution time.

[0077] When a manufacturing node receives tasks from other nodes, the first part of the benefit It is nonlinear, adding a transformation constant variable and constant variable transpose queue Convert task optimization scheduling into a linear problem:

[0078] Among them, w m and is a constant variable, Φ(Ω(t)) represents the maximum comprehensive benefit value, Ω(t) represents the difference between the current comprehensive benefit and the average comprehensive benefit, and E represents the expected value of the benefit. is the average time for task execution;

[0079] The average execution time obtained by maximizing the comprehensive benefits of the collaborative computing system CS and resources Substitute the combination into Φ(Ω(t)) and obtain the optimized task scheduling strategy by minimizing Φ(Ω(t)):

[0080] When the real queue When the queue is not longer than the constant variable transposition queue, the manufacturing node accepts all requests for the mth type of tasks arriving at time t, otherwise it waits.

Claims

1. A task-level collaborative optimization scheduling method supporting multi-cloud network outage disaster recovery, characterized in that: The following steps are involved: Manufacturing nodes assign tasks to other matching manufacturing nodes through a non-service-dependent end-to-end communication channel; The CoopDispatch method is used between manufacturing nodes to coordinate and optimize the scheduling of tasks with other manufacturing nodes; The task processing results of a manufacturing node and other manufacturing nodes are synchronized between the manufacturing nodes through the LocalAny mechanism.

2. According to claim 1, a task-level collaborative optimization scheduling method supporting multi-cloud network failure disaster recovery is characterized in that: The manufacturing node is a client on the edge side of a multi-cloud environment that is used to carry a production control system in a manufacturing enterprise workshop.

3. A task-level collaborative optimization scheduling method supporting multi-cloud network failure disaster recovery according to claim 1 or 2, characterized in that: The multi-cloud includes the cloud formed after the enterprise management and control business is moved to the cloud, and the sub-clouds of the production workshops of various manufacturing enterprises.

4. According to claim 1, a task-level collaborative optimization scheduling method supporting multi-cloud network failure disaster recovery is characterized in that: The CoopDispatch method calculates the minimum average execution time of resource-constrained tasks in a collaborative computing system composed of multiple manufacturing nodes, and collaboratively schedules tasks to improve the overall benefit of the collaborative computing system, including the following steps: 1) The tasks in the collaborative computing system CS are divided into |M| categories, and the task set is represented by Task. For the mth category of tasks, m∈|M|, its tasks are only in the resource R i Task collection m ∈Task, for the current manufacturing node, the arrival rate of the mth task request to other manufacturing nodes at time t satisfy: in, represents the maximum arrival rate of the mth task, i represents the i-th task of the mth task, Is a flag used to indicate resource R i Tasks that can be handled T i Is it in the task set Task m In Chinese, that is: The manufacturing node processes multiple tasks at the same time. The mth task at time t is stored in In the queue, all task queues The following queue matrix Q(t) is constructed: r i Represents a resource set R i A resource, It means processing the tail task of the m-th task whose queue waiting number is |M|; Indicates resource r i The mth type of task processed, i=1…I; The number of task requests of type m received at time t is defined as It satisfies: Among them, Δt represents the time interval and is a constant; The average task execution time is as follows: Among them, α is an empirical constant, E{} represents the expectation, and here represents the average task execution time expect; 2) The collaborative computing system CS allocates the resources required for the task processing task. The resource CS corresponding to the i-th task i Assigned to The resource requested by the mth task is The manufacturing node updates the task processing queue through the following model: in, Indicates the number of tasks of the mth type that can be processed by a unit resource within a set period. 3) Based on the average time of task execution and resources Construct a comprehensive benefit model to maximize the comprehensive benefits of the collaborative computing system CS and obtain the minimum average execution time and resources combination to achieve collaborative optimization scheduling.

5. The task-level collaborative optimization scheduling method supporting multi-cloud network failure disaster recovery according to claim 1 is characterized in that: In step 3), the collaborative optimization scheduling is realized by maximizing the comprehensive benefits of the collaborative computing system CS, including the following steps: (1) When a manufacturing node's task is accepted by other manufacturing nodes, the manufacturing node obtains the following benefits: in, represents the first part of the benefit, β m is the m-th task request benefit constant, is the average time for task execution; (2) The number of task requests of the mth type processed by the manufacturing node at time t, n m (t) is: It indicates the number of tasks of the mth type that can be processed by a unit resource within a set period; The corresponding benefits are: Among them, n m n m (t) time expectation, θ m is the corresponding revenue constant; represents the second part of the benefit, which depends on the number of requests served; (3) The comprehensive benefit model is: Among them, i represents the i-th task of the m-th category of tasks; (4) In order to maximize the comprehensive benefits and achieve collaborative optimization, when B reaches its maximum value, we get is the minimum average execution time.

6. The task-level collaborative optimization scheduling method supporting multi-cloud network failure disaster recovery according to claim 1 is characterized in that: The task processing results of the manufacturing node and other manufacturing nodes are synchronized between the manufacturing nodes through the LocalAny mechanism, including the following steps: When a manufacturing node receives tasks from other nodes, the first part of the benefit It is nonlinear, adding a transformation constant variable And constant variable transpose queue Convert task optimization scheduling into a linear problem: Among them, w m and is a constant variable, Φ(Ω(t)) represents the maximum comprehensive benefit value, Ω(t) represents the difference between the current comprehensive benefit and the average comprehensive benefit, E{} represents the expectation, where it represents the expected value of the benefit, is the average time for task execution; Maximize the average execution time of step 3) by maximizing the comprehensive benefits of the collaborative computing system CS and resources Substitute the combination into Φ(Ω(t)) and obtain the optimized task scheduling strategy by minimizing Φ(Ω(t)): When the real queue When the queue is not longer than the constant variable transposition queue, the manufacturing node accepts all requests for the mth type of tasks arriving at time t, otherwise it waits.

Images