Space-ground integrated network 3c resource scheduling method for mtc task active migration

Abstract

The application discloses a 3C resource scheduling method for MEC task active migration in space-ground integrated network, first, a user directly uploads a task to a LEO satellite node through a direct communication link with the LEO satellite, and simultaneously sends task information and uploaded satellite information to a ground control station; a satellite side MEC node senses the resource state and task state of a neighboring satellite node in real time, determines local processing or migration to other node processing of the task through task priority judgment and sorting of the node, and returns the result from the shortest route to the user side after the satellite node task processing is completed. In the case of limited computing and storage resources of the LEO satellite node, the user task can be processed and successfully delivered as much as possible, the task processing amount is improved, the task processing delay is reduced, the user experience is improved, and the method has the characteristics of high resource utilization, low processing delay, high universality and low complexity.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of satellite communication technology of space-ground integration network, and particularly relates to a 3C resource scheduling method for MEC task active migration of space-ground integration network. BACKGROUND

[0002] With the continuous development of satellite communication technology, it is imperative to build a space-ground integration network, and space-ground integration networking technology is the key to network interconnection and information transmission. Since the current ground network cannot achieve full coverage of global land and sea areas, the network coverage rate in some remote areas (ocean, desert, etc.) is low and the network coverage cost is too high, while satellite networks can effectively supplement the ground network using their high-altitude advantage, thereby achieving higher network coverage. Countries around the world have shifted the focus of communication system construction to the construction of space-ground integration networks.

[0003] In the construction of new low-orbit satellite constellations, the Starlink plan of the American SpaceX company has entered the large-scale intensive deployment stage, the "OneWeb" constellation of the British OneWeb company has launched 3 batches of 74 satellite nodes, and China has launched "Hongyan", "Hongyun", "Tiangxiang" and other verification satellites. A variety of low-orbit satellite constellations that realize or plan to realize satellite-ground networking functions have appeared worldwide. The primary feature of these emerging low-orbit satellite constellations is to achieve ultra-high-speed communication between satellites, which can provide higher quality, low-latency network services to users at a lower cost. However, the existing old central base station maintained communication mode cannot meet the information transmission demand of ultra-high-speed communication, so the satellite edge computing concept is introduced. Satellite edge computing refers to setting up edge computing nodes on communication satellites, making full use of valuable computing and storage resources on the satellite, and placing part of the low-latency task to the satellite edge side for computing and processing. This approach can effectively reduce the number of task backhaul times, reduce the occupation of limited communication bandwidth between satellites and the ground, expand the network coverage range, shorten the service response time, and improve the user experience.

[0004] Although the ultra-high-speed communication capability between satellites provides sufficient bandwidth for edge computing task offloading, the communication resources between satellites and the ground are still limited, and how to effectively deploy and control the satellite constellation edge computing nodes from the ground station is still a difficult problem. In addition, since the edge computing nodes are located in low-orbit satellite constellations, they can carry limited computing and storage resources, and how to fully utilize the limited computing and storage resources to maximize task processing benefits is also a difficult problem that needs to be focused on.

[0005] Traditional low earth orbit satellite (LEO) is independent of the ground communication network, and develops in parallel with the ground network. Due to the limitation of on-board resources, the actual development is slow. After the introduction of the concept of edge computing, a new architecture of LEO constellation and ground network integration is proposed, and the development of space-ground integrated network is also on the road of rapid development. Under the current space-ground integrated network architecture, how to reasonably process and unload the task of low earth orbit satellite edge computing node to realize reasonable resource utilization is a problem to be solved. SUMMARY

[0006] Therefore, the present application provides a 3C resource scheduling method for MEC task active migration in space-ground integrated network, which can process and successfully deliver user tasks as much as possible under the condition of limited satellite node computing and storage resources, and improve the task processing capacity.

[0007] To achieve the above purpose, the technical scheme of the present application includes the following steps:

[0008] Step 1: the user uploads the task directly to the satellite node through the direct communication link with the satellite in the LEO constellation; the satellite extracts the task information and sends the task information and LEO satellite information to the ground control station.

[0009] Step 2: the satellite node side MEC server senses the resource state and task state of the adjacent satellite node in real time, and the MEC server judges and sorts the task priority of the satellite node for the purpose of computing saturation, and decides whether the task is processed locally or migrated to the adjacent satellite node for processing.

[0010] Step 3: after each satellite edge computing node completes task migration and task processing, the task execution result is returned to the ground user according to the shortest route.

[0011] Further, the LEO constellation is composed of a group of satellites, and the LEO constellation is represented as S={S1, S2,..., SM}, S1~SM represents the first to the Mth satellite node; take j={1, 2,..., M}, the satellite Sj carries a computing server with computing capability Cj and storage capability Sj. M},S1~S M is the first to the Mth satellite node; take j={1, 2,..., M}, the satellite Sj carries a computing server with computing capability Cj and storage capability Sj. j j j

[0012] The user's task is T={T1, T2,..., TN}, which is an indivisible task, take i={1, 2,..., N}, the task Ti has a computing size Pi and a required storage space MSi. N i i i ​​​​​​​

[0013] Further, in step two, the MEC server determines, for the purpose of computing saturation, whether the current task can be completed within the time delay requirement. If yes, the current task is processed locally at the satellite node.

[0014] The MEC server determines whether the current task can be completed within the time delay requirement. If yes, the current task is processed locally at the satellite node.

[0015] If the current task cannot be completed within the time delay requirement, and the priority of the current task is higher than the set level, the execution order of the current task is promoted, and a rearranged task execution order list is obtained.

[0016] For the rearranged task execution order list, it is determined whether the current task can be completed within the time delay requirement. If yes, the current task is processed locally at the satellite node. Otherwise, the current task is migrated to a neighboring satellite node for processing.

[0017] Further, in step two, the MEC server determines, for the purpose of computing saturation, whether the current task can be completed within the time delay requirement. If yes, the current task is processed locally at the satellite node.

[0018] An optimization problem is constructed, and the optimization objective is set as the total time delay. For the purpose of computing saturation, the following optimization problem is set:

[0019]

[0020] Where t ij is the processing time delay of the satellite node j executing the task T i ; n j is the number of tasks carried by the jth satellite MEC node; MS i is the storage resource occupied by the task T i ; n jr represents the number of tasks currently being executed by the satellite j; P i is the computation size of the task T i ; C rj and S rj are the remaining computation resource and the remaining storage resource of the current satellite j, respectively.

[0021] The optimization problem is solved, and the task is determined to be processed locally at the satellite node or migrated to a neighboring satellite node for processing, based on the task priority judgment and sorting at the satellite node. The offloading path matrix that satisfies the minimum total time delay and the minimum total time delay are determined.

[0022] Further, the optimization problem is solved, and the following method is used:

[0023] S1: initialize the following parameters:

[0024] task T i the size of the computational load P i , the storage resource MS i occupied by task T i , the highest required delay tddl of task T i , the total storage resource S of satellite node j j , the total computational resource C of satellite node j j , the evaluation parameter α, and the initial task execution order list on satellite j.

[0025] S2: determine and whether the following conditions are met: if yes, go to the next step S3; if not, the satellite j does not have enough resources to complete the current task, and the current task is migrated to a neighboring satellite node for processing.

[0026] S3: for a single task T i on satellite j, calculate whether the completion delay t i of task T ij meets the condition t ij ≤ t ddl , if yes, set the evaluation parameter α of task T i to A, otherwise set the evaluation parameter α of task T i to B.

[0027] For each task on satellite j, perform S3 to obtain its evaluation parameter.

[0028] S4: for a task with an evaluation parameter A and a priority higher than a set level, promote the execution order of the current task, and obtain a rearranged task execution order list; return to S2.

[0029] For a task with an evaluation parameter B, execute the task by satellite j.

[0030] S5: obtain an offloading path matrix according to the task allocation, and calculate the total delay required by the offloading strategy, which is the optimized minimized total delay.

[0031] Preferably, the evaluation parameter α has a value of A=0 and B=∞.

[0032] Beneficial effects:

[0033] The application provides a 3C resource scheduling method for MEC task active migration in a space-ground integrated network, first, a user directly uploads a task to a LEO satellite node through a direct communication link with the LEO satellite, and simultaneously sends task information and uploaded satellite information to a ground control station; a satellite side MEC node senses the resource state and task state of a neighboring satellite node in real time, determines local processing or migration to other node processing of the task through task priority judgment and ordering of the node, and returns the result from the shortest route to the user side after the satellite node task processing is completed. In the case of limited computing and storage resources of the LEO satellite node, the user task can be processed as much as possible and successfully delivered, the task processing amount is improved, the task processing delay is reduced, the user experience is improved, and the method has the characteristics of high resource utilization, low processing delay, high universality and low complexity. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 is an edge computing flowchart of the space-ground integrated network provided by the application.

[0035] Figure 2 is a space-ground integrated network edge computing scene model diagram provided by the application.

[0036] Figure 3 is a satellite side MEC node autonomous scheduling task migration process schematic diagram proposed by the application.

[0037] Figure 4 is an inter-satellite scheduling algorithm flowchart.

[0038] Figure 5 is a comparison diagram of overall edge computing result delivery delay of different algorithms under different task quantities proposed in the embodiment of the application.

[0039] Figure 6 is a comparison diagram of overall edge computing result delivery delay of different algorithms under different satellite quantities proposed in the embodiment of the application.

[0040] Figure 7 is a comparison diagram of inter-satellite edge computing required delay of different algorithms under different task quantities proposed in the embodiment of the application. DETAILED DESCRIPTION

[0041] The application will be described in detail below with reference to the drawings and examples.

[0042] The application provides a 3C resource scheduling method for MEC task active migration in a space-ground integrated network, which comprises the following steps. Figure 1

[0043] Step 1: A user uploads a task directly to a satellite node through a direct communication link with a satellite in a LEO constellation; the satellite extracts task information and sends the task information and LEO satellite information to a ground control station.

[0044] The space-ground integrated network model in the application is shown in Figure 2 The scene mainly comprises a LEO constellation, a ground control station and a user.The LEO constellation comprises a group of satellites S={S1, S2, …, S M}, j={1, 2, …, M}; each satellite S j in the constellation is provided with an edge computing server with a computing capacity of C j and a storage capacity of S j . The ground control station is responsible for the unified deployment of MEC servers in the LEO constellation and the control and decision of the satellites in the constellation that can be currently covered. The user task is T={T1, T2, …, T N}, i={1, 2, …, N}, which is an indivisible task, wherein the computing amount of the task T i is P i , and the required storage space is MS i .

[0045] Step 2: The MEC server on the satellite node side senses the resource state and task state of the adjacent satellite nodes in real time, and judges and sorts the task priority of the satellite node for the purpose of computing saturation, to determine whether the task is processed locally or migrated to an adjacent satellite node for processing.

[0046] The uploading time delay of the user for uploading the task to the satellite side MEC node is wherein d i is the transmission distance between the satellite and the ground, and v j is the communication transmission rate between the satellite and the ground; the returning time delay of the result returned from the satellite side MEC node to the user side is wherein d return is the distance between the returning satellite and the user, and v returnLet be the return transmission rate. In summary, the propagation delay for space-to-ground missions within the space-ground integrated network is t. s =t d +t return The satellite-side MEC server processing latency is... The mission propagation delay for a mission receiving satellite to transfer a mission that cannot be processed in a timely manner to another satellite is t. ISL The queuing delay for lower priority tasks within the satellite-side MEC node is t. q .

[0047] like Figure 4 As shown in the dashed box, step two consists of the following steps:

[0048] The MEC server determines whether the current task can be completed within the latency requirement. If so, the current task is processed locally at this satellite node.

[0049] If the current task cannot be completed within the time limit, and the current task has a higher priority than the set level, then the execution order of the current task is promoted, and a rearranged list of task execution orders is obtained.

[0050] For the rearranged task execution order list, it is further determined whether the current task can be completed within the time delay requirement. If it can, the current task is processed locally at this satellite node; otherwise, the current task is migrated to a neighboring satellite node for processing.

[0051] Furthermore, when scheduling computing tasks across satellites, it's necessary to consider whether the current satellite node has sufficient resources to handle the task locally. In other words, does the receiving satellite node have adequate computing and storage resources to support its completion of the task? Assume the current satellite node's edge server already occupies MC of computing resources. j The occupied storage resources are MS j Therefore, the current S j The satellite server can provide the remaining computing resources (resource C). rj =C j -MC j The remaining storage resources are S rj =S j -MS j To ensure that satellite nodes can process tasks to the maximum extent possible within the limited on-board resources without overloading them, an evaluation parameter α is introduced to assess whether the current satellite node has sufficient resources to complete the current task.

[0052] Therefore, this embodiment of the invention proposes a method to obtain the optimal inter-satellite scheduling solution by constructing an optimization problem. The specific process is as follows:

[0053] Construct an optimization problem, setting the optimization objective as total time delay, and aiming at computational saturation, the optimization problem is as follows:

[0054]

[0055] where t ij is the processing delay of the satellite node j to perform task T i ; n j is the number of tasks carried by the jth satellite MEC node; MS i is the storage resource occupied by task T i ; n jr represents the number of tasks currently being performed by the satellite j; P i is the computation size of task T i ; C rj and S rj are the remaining computation resource and the remaining storage resource of the current satellite j, respectively.

[0056] Solving the optimization problem, in combination with the satellite node task priority judgment and sorting, determines whether the task is processed locally or migrated to a neighboring satellite node for processing, and determines the offloading path matrix that satisfies the minimum total delay and the minimum total delay.

[0057] The storage capacity of the server limits the number of tasks that can be stored on each server, so the satellite side MEC server needs to be scheduled in time to complete the task offloading adjustment and generate the offloading path matrix X r . The above optimization problem is an NP-hard problem and cannot be solved directly, so the present application proposes an improve greedy dispatch algorithm (IGD) algorithm to solve this optimization problem. Specifically, the following method is used:

[0058] S1: initialize the following parameters:

[0059] the computation size P i of task T i , the storage resource MS i occupied by task T i , the highest required delay tddl of task T i , the total storage resource S j of the satellite node j, the total computation resource C j of the satellite node j, the evaluation parameter a, and the initial task execution order list on the satellite j.

[0060] S2: judge whether and are satisfied. If satisfied, go to the next step S3; if not satisfied, the current task is migrated to a neighboring satellite node for processing because the satellite j does not have enough resources to complete the current task.

[0061] S3: for a single task Ti Calculation task T i Completion delay t ij Does t satisfy? ij ≤t ddl If the conditions are met, then task T will be... i The evaluation parameter α is set to A; otherwise, task T is cancelled. i The evaluation parameter α is set to B;

[0062] For each mission on satellite j, S3 is executed to obtain its evaluation parameters.

[0063] S4: For a task with an evaluation parameter of A and a priority higher than the set level, improve the execution order of the current task and obtain the rearranged task execution order list; return to S2.

[0064] For a task with evaluation parameter B, satellite j will perform the task.

[0065] S5: Obtain the unloading path matrix based on the task allocation, and calculate the total latency required by the unloading strategy, which is the optimized and minimized total latency.

[0066] In this embodiment of the invention, the evaluation parameter α is set to 0 for A and ∞ for B.

[0067] Step 3: After each satellite edge computing node completes task migration and task processing, it returns the task execution results to the ground user according to the shortest route.

[0068] Example

[0069] The invention is designed as follows Figure 2 The image shows a space-ground integrated network edge computing scenario. By adjusting the number of tasks and satellites, the IGD algorithm of this invention and other common offloading strategy algorithms are applied to this scenario. The overall latency required for edge computing and the inter-satellite latency are compared. Each experimental result is obtained after repeating the experiment 300 times.

[0070] Figure 5The simulation comparison of the overall time delay required for returning results of different algorithms under different task quantities is carried out, and the satellite quantity is M=8. It can be known from the analysis result of the figure that the latest IGD algorithm proposed in the application has the advantage of shorter result return time delay compared with the other two algorithms under a certain satellite quantity, and this advantage is particularly obvious when the task quantity increases. The time delay of the traditional greedy algorithm and the random allocation algorithm increases greatly when the task quantity is high, and the trend is similar, wherein the traditional greedy algorithm shows superior performance. The simulation result of the comparison with the other two algorithms, especially when the task quantity is high, shows that the inter-satellite scheduling IGD algorithm proposed in the application can help users to receive the edge computing processing result of the satellite as soon as possible under a certain satellite quantity, effectively shorten the task processing time delay, and show more superior performance.

[0071] Figure 6 The simulation result comparison of the influence of different satellite quantities on the task result return time under a certain task quantity is carried out, and the task quantity is N=4. It can be easily obtained from the result that the performance of each allocation strategy algorithm is consistent when the satellite quantity is 1, and the IGD algorithm quickly shows higher task processing efficiency when the satellite quantity gradually increases, and the time delay required for result return is obviously shorter than that of the other two algorithms.

[0072] Figure 7 The simulation comparison of the inter-satellite task processing time of different algorithms under a certain satellite quantity is carried out, and the satellite quantity is M=6. It can be obtained from the simulation result analysis that the inter-satellite processing time delay of the IGD algorithm is shorter than that of the other two algorithms under the same satellite quantity and task quantity, which shows that the inter-satellite scheduling algorithm in the IGD proposed in the application has an absolute advantage in the migration and scheduling of inter-satellite tasks compared with the other two algorithms. And due to the instability of the random allocation algorithm, the inter-satellite resources are wasted, which causes higher inter-satellite task processing time delay.

[0073] In summary, the above is only a preferred embodiment of the application, and is not used to limit the protection scope of the application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the protection scope of the application.

Claims

1.A 3C resource scheduling method for MEC task active migration in space-ground integrated network, characterized in that, Comprising the following steps: Step one: the user uploads the task directly to the satellite node through the direct communication link with the satellite in the LEO constellation; the satellite extracts the task information and sends the task information to the ground control station with the LEO satellite information; the LEO constellation is composed of a group of satellites, and the LEO constellation is represented as S = {S1, S2, …, S M}, S1~S M are the first to the M-th satellite nodes; take j = {1, 2, …, M}, the satellite S j carries an edge computing server with a computing capacity of C j and a storage capacity of S j ; The user's task is T = {T1, T2, ..., T} N Let} be an indivisible task, with i = {1, 2, ..., N}, and task T. i The computational complexity is P i The required storage space is MS i ; Step two: the satellite node side MEC server perceives the resource state and task state of the adjacent satellite nodes in real time, and the MEC server determines the task priority of the satellite node and sorts it for the purpose of calculation saturation, decides the task to be processed locally or migrated to the adjacent satellite node for processing, and the specific process is: An optimization problem is constructed, the optimization objective is set as the total delay, and the following optimization problem is set for the purpose of calculation saturation: where t ij is the processing delay of the satellite node j to perform the task T i ; n j is the number of tasks carried by the jth satellite MEC node. MS i for task T i occupies storage resources; n jr denotes the number of tasks currently being executed by satellite j; P i is the computation size of task T i ; C rj and S rj are the remaining computation resource and remaining storage resource of current satellite j, respectively. Solving the optimization problem, combining the task priority judgment and sorting of the satellite node, deciding the task to be processed locally or migrated to the adjacent satellite node for processing, determining the unloading path matrix meeting the minimum total delay and the minimum total delay corresponding to the minimum total delay; Step three: after each satellite edge computing node completes task migration and task processing, the task execution result is returned to the ground user according to the shortest route; In the step two, the MEC server determines the task priority of the satellite node and sorts it for the purpose of calculation saturation, decides the task to be processed locally or migrated to the adjacent satellite node for processing, and the specific process is: The MEC server judges whether the current task can be completed within the time delay requirement, if yes, the current task is processed locally at the satellite node; If the current task cannot be completed within the time delay requirement, and the priority of the current task is higher than the set level, the execution order of the current task is promoted, and the rearranged task execution order list is obtained; For the rearranged task execution order list, continue to judge whether the current task can be completed within the time delay requirement, if yes, the current task is processed locally at the satellite node, otherwise the current task is migrated to the adjacent satellite node for processing. 2.The MEC task-oriented active migration terrestrial-satellite integrated network 3C resource scheduling method of claim 1, wherein, The solving of the optimization problem is specifically as follows: S1: initialize the following parameters: Task T i Size P of the amount of computation i Task T i Occupied storage resource MS i Task T i Highest required latency t ddl Total storage resource S of satellite node j j Total computation resource C of satellite node j j Evaluation parameter a and initial task execution order list on satellite j S2: judging with if satisfied, go to next step S3; if not satisfied, satellite j has not enough resources to complete the current task, the current task is migrated to the neighboring satellite node for processing; S3: for a single task T on satellite j i , compute the task T i completion delay t ij ; check if t ij ≤ t ddl ; if yes, set the task T i evaluation parameter a to A, otherwise set the task T i evaluation parameter a to B; S3 is performed for each task on satellite j to obtain its evaluation parameter; S4: for the task with evaluation parameter A and priority higher than the set level, the execution order of the current task is promoted, and the rearranged task execution order list is obtained; Return to S2; For the task with evaluation parameter B, the satellite j executes the task; S5: according to the task allocation, the unloading path matrix is obtained, and the total delay required for unloading is the minimum total delay after optimization. 3.The method of claim 2, wherein, The value of the evaluation parameter a is 0, and the value of B is infinity.

Images