Task scheduling method and system based on service deployment update

Abstract

The application relates to a service deployment update-based task scheduling method and system, which comprises the following steps: S1, initializing service deployment of each edge server in an edge network; S2, discretizing system time into a plurality of time frames, each time frame being composed of a plurality of time slots, service deployment occurring in the first time slot of each time frame, and task scheduling occurring in any time slot after the first time slot of each time frame; S3, at the beginning of each time frame, dynamically adjusting service deployment to obtain target service deployment corresponding to target requirements of timeout loss of a task set; S4, based on the target service deployment, calculating scheduling priorities of tasks of the task set by the edge server; and S5, based on the scheduling priorities of the tasks, scheduling the tasks. The application separates service deployment and task scheduling in time, dynamically adjusts service deployment, and determines a scheduling strategy according to priorities in task scheduling, thereby improving the service quality of users.

Description

Technical Field

[0001] This invention belongs to the field of edge computing technology, specifically relating to a task scheduling method and system based on service deployment and updates. Background Technology

[0002] The need to process massive amounts of data has spurred the development of numerous data-intensive applications. These applications operate with data at their core, continuously acquiring, updating, and storing large amounts of data, and performing various calculations on this data. Traditional cloud computing technology, based on the powerful computing and storage capabilities of data centers, can handle the vast amounts of data collected from terminal devices. However, with the development of data-intensive applications, users increasingly demand a smooth user experience and have higher requirements for task response time and service quality, making it difficult for cloud computing to meet the needs of many application scenarios.

[0003] Mobile Edge Computing (MEC) reduces network latency, a significant portion of traditional cloud computing, by deploying computing and storage resources—originally belonging to cloud data centers—to the edge of the wireless access network. This provides end users with computing and storage capabilities while reducing energy consumption. Unlike traditional task offloading, MEC, due to the dependency of user requests, can only process user tasks for a certain type of service after the edge server has deployed application-related service caches. Limited by storage resources, edge servers can only selectively deploy some services, and different deployment schemes can influence task offloading decisions. Furthermore, from the perspective of the edge cluster, using efficient task scheduling algorithms to provide high-quality services to users is crucial. Considering that some tasks have response latency constraints, reducing the response time of user tasks to meet the needs of latency-sensitive tasks becomes key to ensuring user service quality. Inappropriate task scheduling methods may overload some edge nodes, leading to a decrease in the quality of service provided.

[0004] Service deployment and task scheduling are tightly coupled; the service deployment strategy determines the decision space for task scheduling, while the task scheduling result reflects the performance of the service deployment strategy. Existing task scheduling methods lack a comprehensive consideration of service deployment. Summary of the Invention

[0005] In view of the above-mentioned shortcomings of the existing technology, the purpose of the present invention is to provide a task scheduling method and system based on service deployment and updates.

[0006] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0007] The task scheduling method based on service deployment updates includes the following steps:

[0008] S1. Initialize the service deployment of each edge server in the edge network;

[0009] S2. Discretize the system time into several time frames, each time frame consists of several time slots. Service deployment occurs in the first time slot of each time frame, and task scheduling occurs in any time slot after the first time slot of each time frame.

[0010] S3. At the beginning of each time frame, dynamically adjust the service deployment to achieve the target service deployment corresponding to the timeout loss of the task set.

[0011] S4. Based on the target service deployment, calculate the scheduling priority of tasks in the task set for the edge server;

[0012] S5. Perform task scheduling based on task scheduling priority.

[0013] As a preferred embodiment, step S1 includes:

[0014] Sort all edge servers within the edge network according to their storage capacity, and divide the set of edge servers into high-capacity groups N based on a first storage capacity threshold. high and low-capacity group N low ;

[0015] The application set corresponding to the service is sorted according to the storage capacity required by the application, and the application set is divided into high-demand group S according to the second storage capacity partitioning threshold. high and low demand group S low ;

[0016] According to the Top-K algorithm, S high Applications deployed to N high Each edge server in the S low Applications deployed to N low The edge servers in the system.

[0017] As a preferred embodiment, in step S3, the timeout loss for the task set is:

[0018]

[0019] T u t represents the response time for task u in the task set to be unloaded to its corresponding edge server. u w represents the expected response time of task u in the task set. u This represents the weighting coefficient, used to measure the sensitivity of task u to latency. Its magnitude is related to the service type corresponding to the task.

[0020] As a preferred option, the response time for task u to be unloaded to its corresponding edge server n is:

[0021]

[0022] in, The transmission latency for sending task data from task u to edge server n. Queuing latency for executing task u on edge server n, The computation latency for executing task u on edge server n.

[0023] As a preferred embodiment, the target requirement in step S3 is a preset timeout loss threshold.

[0024] As a preferred embodiment, step S4 includes:

[0025] For task u, if it is scheduled to edge server n, and let the position of task u in the waiting queue of edge server n be idx, then the start time of task u is the completion time of its predecessor task; combined with the computational complexity c of task u... u Computing power W of edge server n n Then the estimated response time for task u is:

[0026]

[0027] Where curTime represents the current time, and arrivalTime represents the arrival time. u Indicates the time when task u was generated. To calculate the delay, To account for queuing delay; due to the addition of task u, the tasks in the waiting queue that are after task u have a delayed response time compared to the previous actual response time, which will cause tasks that could have been completed within the expected response time to fail to complete on time. An increment Δ is used to account for the timeout loss. u,n To describe the change in timeout loss of the task set due to the addition of a new task;

[0028] For each task v that follows the new task u, its estimated response time needs to be recalculated. Combined with the previously estimated response time for this task Obtain the timeout loss increment Δ u,n The specific calculation method is as follows.

[0029]

[0030] Where qsize represents the length of the task queue of edge server n;

[0031] The estimated response time T for task u on edge server n is obtained. u,nAnd the timeout penalty increment Δ for task u on edge server n. u,n Then, the scheduling priority of edge server n for task u is calculated as follows:

[0032]

[0033] Where α1 is the task scheduling weight coefficient, and β is the scaling factor.

[0034] As a preferred option, the scheduling priority E n,u The smaller the value, the higher the priority of edge server n when it is selected as the target server for task scheduling when facing task u.

[0035] As a preferred option, the edge server's computing power is equivalent to CPU processing power.

[0036] This invention also provides a task scheduling system based on service deployment updates, applying the task scheduling method described in any of the preceding solutions, wherein the task scheduling system includes:

[0037] The service deployment initialization module is used to initialize the service deployment of each edge server in the edge network;

[0038] The time discretization module is used to discretize the system time into several time frames. Each time frame consists of several time slots. Service deployment occurs in the first time slot of each time frame, and task scheduling occurs in any time slot after the first time slot of each time frame.

[0039] The service deployment adjustment module is used to dynamically adjust the service deployment at the beginning of each time frame to achieve the target service deployment corresponding to the timeout loss of the task set.

[0040] The priority calculation module is used to calculate the scheduling priority of tasks on the task set for the edge server based on the target service deployment.

[0041] The task scheduling module is used to schedule tasks based on their scheduling priorities.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] This invention divides system time into several time slots, separating service deployment and task scheduling. A local search algorithm is used to adjust the service deployment scheme, and during task scheduling, the priority of nodes is calculated based on the current service deployment scheme to determine which node to schedule the task for execution, thereby improving the user's service quality. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of the model components in an embodiment of the present invention;

[0045] Figure 2 This is a schematic diagram of the system time scale division according to an embodiment of the present invention;

[0046] Figure 3 This is a schematic diagram illustrating the dynamic adjustment of service deployment according to an embodiment of the present invention;

[0047] Figure 4 This is a schematic diagram of user request scheduling according to an embodiment of the present invention;

[0048] Figure 5 This is an execution flowchart of the task scheduling method according to an embodiment of the present invention. Detailed Implementation

[0049] To more clearly illustrate the embodiments of the present invention, specific implementation methods will be described below with reference to the accompanying drawings. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings and other implementation methods can be obtained based on these drawings without any creative effort.

[0050] The task scheduling method based on service deployment and updates according to embodiments of the present invention, such as Figure 1 As shown, this method mainly includes three components: 1) User service quality quantification module; 2) Service deployment module; 3) Task scheduling module.

[0051] (I) User Service Quality Quantification Module

[0052] Each task u can use <d u ,c u ,t u ,w u > to describe, where d u Indicates the amount of data carried in this request; c u This indicates the computational cost required to process the request; t u This indicates the expected response time for a user's request. Tasks that fail to receive a result within the expected response time will result in a decrease in service quality. u This represents a weighting coefficient used to measure the sensitivity of a user request to latency. Its magnitude is related to the type of service the request corresponds to; for services with higher real-time requirements, w... u The value is relatively larger.

[0053] The key resources of edge servers lie in computing power and storage capacity. For any edge server n, its resource characteristics can be abstracted as follows: <W n ,R n >, where W n and R n These represent the CPU processing power and storage capacity of edge server n, respectively.

[0054] From the perspective of edge service platforms, the key to providing high service quality lies in whether the processing result of task u can be delivered within its expected response time T. u Internally relayed back to the user. The response time for a user request (i.e., the task) consists of three parts:

[0055] (1) Transmission latency of user sending task data to the target edge server

[0056] (2) Processing latency of MEC servers (i.e., edge servers) in executing tasks, including queuing latency. and computational delay

[0057] (3) Transmission delay for users to receive the processing results from the MEC server.

[0058] Considering that the data volume of task processing results is usually relatively small, the transmission latency for users to receive processing results from the MEC server is ignored.

[0059] Therefore, the response time T for user task u to be offloaded to edge server n u,n It can be represented as

[0060] However, when the user request density exceeds the processing capacity of the edge service platform, the actual response time of some tasks will exceed the expected response time, resulting in a decline in service quality. This invention defines timeout loss as a measure of user service quality; the smaller the timeout loss of a task set, the higher the relative service quality. The definition of timeout loss is as follows:

[0061]

[0062] (II) Service Deployment Module

[0063] like Figure 2 As shown, this invention separates the time scales of service deployment and task scheduling decisions to reduce the frequency of service deployment adjustments. System time is discretized into several time slots, with each T0 time slot constituting a time frame. Service deployment occurs at the beginning of each time frame to avoid high service deployment costs; task scheduling occurs in any intermediate time slot to maintain real-time response. Within the time scale of a single time frame, all task scheduling decisions are based on the service deployment decisions obtained at the beginning of each time frame.

[0064] Initially, service initialization deployment needs to be performed on each edge server. The entire service initialization deployment is divided into two parts: resource initialization and service pre-deployment. First, the resource initialization process sorts all servers in the edge network in descending order of storage capacity, and divides the server set N into N partitions according to the MEC resource partitioning threshold θ1. high and N low Two groups, where N high The MEC server in the middle has relatively abundant storage resources, N low The MEC server in the system has relatively limited storage resources. After resource initialization is complete, service pre-deployment begins. First, the application set is divided into S groups based on the storage capacity threshold θ2 required by each application. high and S low Two groups. Under the constraint of storage resources, traverse N... high The edge servers in the network are ranked according to the Top-K algorithm. high The application is deployed to each edge server; following the same method, S low Applications deployed to N low On each edge server in the network.

[0065] After completing the service initialization deployment, the task scheduling strategy is executed and the current timeout loss is recorded. At the beginning of each time frame, a local service deployment adjustment is performed. If an operation optimizes the timeout loss of the task set by at least (1-ε) (ε is a constant in the range (0,1], which ensures the stability and convergence of the values), then the local adjustment is accepted and a new round of iteration begins. If the optimization requirement is not met, the adjustment operation is rejected. If all possible operations fail to optimize the timeout loss of the task set, it indicates that the current service deployment strategy has reached a local minimum, and the service deployment adjustment ends.

[0066] like Figure 3 The diagram shown illustrates the dynamic adjustment of services.

[0067] (III) Task Scheduling Module

[0068] Task scheduling decisions occur in the middle time slot of each time frame and require consideration of the deployment plan at the beginning of the time frame. The scheduling decision aims to minimize timeout penalties. The timeout penalty of a task set is influenced by two factors: firstly, when the actual response time of a task exceeds the expected response time, it increases the overall timeout penalty of the task set; secondly, if a task has a higher timeout penalty coefficient, the timeout penalty after that task will be higher than that of tasks with lower timeout penalty coefficients. Based on this, the concept of node scheduling priority is proposed, and a task scheduling algorithm is designed based on node priority.

[0069] For a given task u, if it is scheduled to edge server n, assuming its position in the server's waiting queue is idx, then task u's start processing time is equal to the completion time of its preceding task. Considering the computational cost c of task u... u Computing power W of edge server n n The estimated response time T for task u is given below. u,n The calculation method is as follows, where curTime represents the current time, and arrivalTime... u Indicates the time when task u was generated.

[0070]

[0071] Because of the addition of task u, the tasks in the cache queue that are after u have a delayed response time compared to the previous actual response time. This may cause tasks that could have been completed within the expected response time to fail to complete on time. Therefore, a timeout penalty increment Δ is needed. u,n This describes the change in task set timeout loss caused by adding a new task. For each task v that follows the new task u, we need to recalculate its estimated response time. Combined with the previously estimated response time for this task The timeout loss increment Δ can be obtained. u,n The specific calculation method is as follows, where qsize represents the task queue length of edge server n.

[0072]

[0073] The estimated response time T for task u on edge server n is obtained. u,n And the timeout penalty increment Δ for task u on edge server n. u,n Then, the scheduling priority of the edge server can be calculated, as follows:

[0074]

[0075] Among them, E n,u E represents the scheduling priority of node n when facing user task u. n,u The smaller the value, the higher the priority of node n when facing task u as the target server for task scheduling, meaning it is more likely to schedule the task to that node. α1 is the task scheduling weight coefficient; β is a scaling factor used to control the timeout loss increment score within a reasonable range, avoiding weight skew caused by an excessive difference between the timeout loss increment score and the estimated response time score of the task.

[0076] The following is an algorithm description of the task scheduling module.

[0077]

[0078]

[0079] like Figure 4 The diagram shown illustrates the user request scheduling process.

[0080] The specific execution steps of this invention are as follows:

[0081] Step 1: The service deployment module discretizes the system time into several time frames, each time frame consisting of several time slots. Service deployment occurs in the first time slot of each time frame, and task scheduling occurs in any intermediate time slot.

[0082] Step 2: The service deployment module divides the server and user requests into two groups based on the amount of resources they have and the amount of resources they require, and then performs pre-deployment of the services.

[0083] Step 3: At the beginning of each time frame, the service deployment module is responsible for dynamically adjusting the service deployment. It enumerates all adjustment schemes, calculates the impact of the adjustments on the overall system service quality using the service quality quantification module, and selects the adjustment scheme that improves service quality.

[0084] Step 4: In the middle time slot of each time frame, the task scheduling module is responsible for scheduling user requests. First, it needs to calculate the estimated response time of the task based on the computational load of the task and the computing power of the edge server.

[0085] Step 5: The task scheduling module calculates the change in system service quality caused by adding a user request to a certain edge service through the service quality quantification module.

[0086] Step 6: Calculate the priority for each edge server based on the results of steps 4 and 5. Then, schedule user requests to the edge node with the highest priority for execution.

[0087] Step 7: Repeat steps 3 through 6.

[0088] like Figure 5 This is a flowchart illustrating the task scheduling method according to an embodiment of the present invention, specifically including the following steps:

[0089] S1. Initialize the service deployment of each edge server in the edge network.

[0090] Specifically, the initialization process includes:

[0091] Sort all edge servers within the edge network according to their storage capacity, and divide the set of edge servers into high-capacity groups N based on a first storage capacity threshold. high and low-capacity group Nlow ;

[0092] The application set corresponding to the service is sorted according to the storage capacity required by the application, and the application set is divided into high-demand group S according to the second storage capacity partitioning threshold. high and low demand group S low ;

[0093] According to the Top-K algorithm, S high Applications deployed to N high Each edge server in the S low Applications deployed to N low The edge servers in the system.

[0094] S2. Discretize the system time into several time frames, each time frame consisting of several time slots. Service deployment occurs in the first time slot of each time frame, and task scheduling occurs in any time slot after the first time slot of each time frame.

[0095] S3. At the beginning of each time frame, the service deployment is dynamically adjusted to achieve the target service deployment corresponding to the timeout loss of the task set.

[0096] The timeout penalty for the task set is as follows:

[0097]

[0098] T u t represents the response time for task u in the task set to be unloaded to its corresponding edge server. u w represents the expected response time of task u in the task set. u This represents the weighting coefficient, used to measure the sensitivity of task u to latency. Its magnitude is related to the service type corresponding to the task.

[0099] The objective requirement of this embodiment of the invention is a preset timeout loss threshold.

[0100] S4. Based on the target service deployment, calculate the scheduling priority of tasks in the task set for the edge server.

[0101] Specifically, for task u, if it is scheduled to edge server n, and let the position of task u in the waiting queue of edge server n be idx, then the start time of task u is the completion time of its predecessor task; combined with the computational complexity c of task u... u Computing power W of edge server n n Then the estimated response time for task u is:

[0102]

[0103] Where curTime represents the current time, and arrivalTime represents the arrival time. u Indicates the time when task u was generated. To calculate the delay, The queuing delay is explained by the fact that due to the addition of task u, the tasks in the waiting queue that are after task u have a delayed response time compared to the previous actual response time. This will cause tasks that could have been completed within the expected response time to fail to complete on time. An increment Δ is used to account for the timeout. u,n To describe the change in timeout loss of the task set due to the addition of a new task;

[0104] For each task v that follows the new task u, its estimated response time needs to be recalculated. Combined with the previously estimated response time for this task Obtain the timeout loss increment Δ u,n The specific calculation method is as follows.

[0105]

[0106] Where qsize represents the length of the task queue of edge server n;

[0107] The estimated response time T for task u on edge server n is obtained. u,n And the timeout penalty increment Δ for task u on edge server n. u,n Then, the scheduling priority of edge server n for task u is calculated as follows:

[0108]

[0109] Where α1 is the task scheduling weight coefficient, and β is the scaling factor.

[0110] Scheduling priority E n,u The smaller the value, the higher the priority of edge server n when it is selected as the target server for task scheduling when facing task u.

[0111] The computing power of the edge server in this embodiment of the invention is CPU processing power.

[0112] S5. Perform task scheduling based on task scheduling priority.

[0113] Correspondingly, embodiments of the present invention also provide a task scheduling system based on service deployment updates, applying the task scheduling method described above in embodiments of the present invention.

[0114] The task scheduling system of this invention includes a service deployment initialization module, a time discretization module, a service deployment adjustment module, a priority calculation module, and a task scheduling module.

[0115] The service deployment initialization module of this invention is used to initialize the service deployment of each edge server in the edge network.

[0116] The initialization process includes:

[0117] Sort all edge servers within the edge network according to their storage capacity, and divide the set of edge servers into high-capacity groups N based on a first storage capacity threshold. high and low-capacity group N low ;

[0118] The application set corresponding to the service is sorted according to the storage capacity required by the application, and the application set is divided into high-demand group S according to the second storage capacity partitioning threshold. high and low demand group S low ;

[0119] According to the Top-K algorithm, S high Applications deployed to N high Each edge server in the S low Applications deployed to N low The edge servers in the system.

[0120] The time discretization module of this invention is used to discretize the system time into several time frames, each time frame consisting of several time slots. Service deployment occurs in the first time slot of each time frame, and task scheduling occurs in any time slot after the first time slot of each time frame.

[0121] The service deployment adjustment module in this embodiment is used to dynamically adjust the service deployment at the beginning of each time frame to achieve the target service deployment corresponding to the timeout loss of the task set meeting the target requirements.

[0122] The timeout penalty for the task set is as follows:

[0123]

[0124] T u t represents the response time for task u in the task set to be unloaded to its corresponding edge server. u w represents the expected response time of task u in the task set. u This represents the weighting coefficient, used to measure the sensitivity of task u to latency. Its magnitude is related to the service type corresponding to the task.

[0125] The objective requirement of this embodiment of the invention is a preset timeout loss threshold.

[0126] The priority calculation module in this embodiment of the invention is used to calculate the scheduling priority of tasks in the task set by the edge server based on the target service deployment.

[0127] Specifically, for task u, if it is scheduled to edge server n, and let the position of task u in the waiting queue of edge server n be idx, then the start time of task u is the completion time of its predecessor task; combined with the computational complexity c of task u... u Computing power W of edge server n n Then the estimated response time for task u is:

[0128]

[0129] Where curTime represents the current time, and arrivalTime represents the arrival time. u Indicates the time when task u was generated. To calculate the delay, The queuing delay is explained by the fact that due to the addition of task u, the tasks in the waiting queue that are after task u have a delayed response time compared to the previous actual response time. This will cause tasks that could have been completed within the expected response time to fail to complete on time. An increment Δ is used to account for the timeout. u,n To describe the change in timeout loss of the task set due to the addition of a new task;

[0130] For each task v that follows the new task u, its estimated response time needs to be recalculated. Combined with the previously estimated response time for this task Obtain the timeout loss increment Δ u,n The specific calculation method is as follows.

[0131]

[0132] Where qsize represents the length of the task queue of edge server n;

[0133] The estimated response time T for task u on edge server n is obtained. u,n And the timeout penalty increment Δ for task u on edge server n. u,n Then, the scheduling priority of edge server n for task u is calculated as follows:

[0134]

[0135] Where α1 is the task scheduling weight coefficient, and β is the scaling factor.

[0136] Scheduling priority E n,u The smaller the value, the higher the priority of edge server n when it is selected as the target server for task scheduling when facing task u.

[0137] The computing power of the edge server in this embodiment of the invention is CPU processing power.

[0138] The task scheduling module of this invention is used to schedule tasks based on their scheduling priorities.

[0139] The above description is merely a detailed explanation of preferred embodiments and principles of the present invention. For those skilled in the art, there may be changes in specific implementation methods based on the ideas provided by the present invention, and these changes should also be considered within the scope of protection of the present invention.

Claims

1. A task scheduling method based on service deployment and updates, characterized in that, Includes the following steps: S1. Initialize the service deployment of each edge server in the edge network; S2. Discretize the system time into several time frames, each time frame consists of several time slots. Service deployment occurs in the first time slot of each time frame, and task scheduling occurs in any time slot after the first time slot of each time frame. S3. At the beginning of each time frame, dynamically adjust the service deployment to achieve the target service deployment corresponding to the timeout loss of the task set. S4. Based on the target service deployment, calculate the scheduling priority of tasks in the task set for the edge server; S5. Perform task scheduling based on task scheduling priority; Step S1 includes: All edge servers within the edge network are sorted by storage capacity, and the set of edge servers is divided into high-capacity groups based on a first storage capacity threshold. and low-capacity groups ; The application set corresponding to the service is sorted according to the storage capacity required by the application, and the application set is divided into high-demand groups according to the second storage capacity partitioning threshold. and low demand group ; According to the Top-K algorithm Applications deployed to The edge servers in the middle, will Applications deployed to Each edge server in the process; In step S3, the timeout loss for the task set is: For tasks in the task set The response time of unloading to its corresponding edge server. For tasks in the task set u Expected response time This represents the weighting coefficient, used to measure the task. u The sensitivity to latency is related to the type of service corresponding to the task; Task Unload to its corresponding edge server n The response time is: in, To send tasks Task data to edge server n Transmission delay, For edge servers n Execute the task u Queuing delay, For edge servers n Execute the task u The computational delay; In step S3, the target requirement is a preset timeout loss threshold. Step S4 includes: For the task If it is scheduled to the edge server Set a task On the edge server The position in the waiting queue is Then the task The start time of processing is the completion time of its previous task; combined with the task computational load With edge servers computing power Then the task The estimated response time is: in, Indicates the current moment. Indicates task The time of its creation, To calculate the delay, Due to queuing delay; due to task The addition of [the task] to the waiting queue Subsequent tasks, with their actual response times delayed compared to previous ones, may cause tasks that were originally expected to be completed within the expected response time to fail to be completed on time. An incremental timeout penalty will be applied. To describe the change in timeout loss of the task set due to the addition of a new task; For each new task The next task It is necessary to recalculate its estimated response time. Combined with the previously estimated response time for this task Receive the timeout loss increment for: in, Represents edge server Task queue length; After receiving the task On the edge server Estimated response time and tasks On the edge server Incremental timeout loss Then, compute edge servers n The scheduling priority for tasks is as follows: in, 1 represents the task scheduling weight coefficient. As a scaling factor; The scheduling priority The smaller the edge server n Facing the task The higher the priority of the target server selected for task scheduling.

2. The task scheduling method based on service deployment updates according to claim 1, characterized in that, The computing power of an edge server is equivalent to CPU processing power.

3. A task scheduling system based on service deployment and updates, using the task scheduling method as described in any one of claims 1-2, characterized in that, The task scheduling system includes: The service deployment initialization module is used to initialize the service deployment of each edge server in the edge network; The time discretization module is used to discretize the system time into several time frames. Each time frame consists of several time slots. Service deployment occurs in the first time slot of each time frame, and task scheduling occurs in any time slot after the first time slot of each time frame. The service deployment adjustment module is used to dynamically adjust the service deployment at the beginning of each time frame to achieve the target service deployment corresponding to the timeout loss of the task set. The priority calculation module is used to calculate the scheduling priority of tasks on the task set for the edge server based on the target service deployment. The task scheduling module is used to schedule tasks based on their scheduling priorities.

Images