Edge Computing Task Scheduling Method and System in Industrial Internet

Abstract

The present invention discloses a method and system for edge computing task scheduling in the industrial Internet. The method includes: obtaining a computing task; searching in the cache of the edge server to see if there is a computing result matching the computing task. If there is, the edge server returns the computing result; if there is no such computing result in the cache of the edge server, search in the blockchain; if there is a computing result in the blockchain, select a scheduling location according to the minimum cost; if not, select a scheduling location according to latency and resources; based on the cache pool and the blockchain cache, establish a scheduling model according to the information of the computing task; solve the scheduling model to determine whether the computing task is scheduled to the edge server. By establishing a joint optimization model of task caching, blockchain, and task scheduling, the present invention combines blockchain technology with the three-layer architecture of industrial devices, edge computing, and cloud computing, reducing task consumption while improving data security during the task scheduling process.

Description

Technical Field

[0001] The present invention relates to the field of industrial Internet technology, and in particular to a method and system for scheduling edge computing tasks in the industrial Internet. Background Art

[0002] Industrial equipment is constrained by limited resources, such as CPU power, storage space, and energy capacity. This makes complex computing tasks inefficient when performed locally on the device. To address this issue, an effective solution is edge computing. Edge computing dispatches computing tasks from local devices to more powerful edge servers, leveraging the edge server's robust computing power to accelerate task processing. Edge servers can pre-process data from different types of devices, facilitating communication between them. Compared to existing cloud computing frameworks, edge computing, with its high bandwidth and low latency, can effectively meet the computing resource demands of industrial production.

[0003] In the Industrial Internet, some computing tasks are highly similar. These repetitive and unordered tasks increase the computational burden on edge servers. Using cache to store dynamic data in edge computing can effectively reduce computing latency and energy consumption. By caching the input or output of computing tasks, when tasks are highly similar or consistent, data in the cache pool is used to reduce resource reuse and data transmission frequency.

[0004] However, in the Industrial Internet, the distributed and highly heterogeneous nature of industrial equipment leads to trust issues between edge servers. Furthermore, as requests for data transmission through channels increase, failure to promptly schedule and monitor policy servers can easily lead to uncontrollable and unreliable data, such as confusion, redundancy, and loss. This can also increase the risk of data leakage and tampering. Failure to address data security issues in task scheduling will not only fail to reduce task processing time, but will also compromise factory data security. Summary of the Invention

[0005] Embodiments of the present invention provide an edge computing task scheduling method and system in the industrial Internet, which are used to solve the problem of repeated calculation of similar tasks and the problem of uncontrollable data caused by distrust between edge servers in existing task scheduling technologies.

[0006] On the one hand, an embodiment of the present invention provides a method for scheduling edge computing tasks in the industrial Internet, including:

[0007] Get computing tasks;

[0008] Search in the cache of the edge server to check if there is a calculation result that matches the calculation task. If there is, the edge server returns the calculation result;

[0009] If there is no matching calculation result in the cache of the edge server, search in the blockchain to check if there is a calculation result that matches the calculation task;

[0010] If there is a calculation result in the blockchain, select the scheduling location according to the minimum cost; if not, select the scheduling location according to the latency and resources, and at the same time update the calculation result in the edge cache pool and the blockchain cache to obtain a reward;

[0011] Based on the edge cache pool and the blockchain cache, establish a scheduling model according to the information of the calculation task;

[0012] Use the task scheduling strategy algorithm based on the improved genetic algorithm to solve the scheduling model and determine whether the calculation task is scheduled to the edge server.

[0013] On the other hand, the embodiment of the present invention provides an edge computing task scheduling system in the industrial Internet, including:

[0014] A policy server, which is used to obtain the calculation task and search in the cache of the edge server to check if there is a calculation result that matches the calculation task. If there is, the edge server returns the calculation result. If there is no matching calculation result in the cache of the edge server, the policy server searches in the blockchain to check if there is a calculation result that matches the calculation task;

[0015] If there is a calculation result in the blockchain, the policy server selects the scheduling location according to the minimum cost; if not, the policy server selects the scheduling location according to the latency and resources, and at the same time updates the calculation result in the edge cache pool and the blockchain cache to obtain a reward;

[0016] Based on the edge cache pool and the blockchain cache, the policy server establishes a scheduling model according to the information of the calculation task, and then uses the task scheduling strategy algorithm based on the improved genetic algorithm to solve the scheduling model to determine whether the calculation task is scheduled to the edge server;

[0017] Industrial equipment executes industrial computing tasks, which are used to schedule the calculation tasks to be executed to the edge server for execution;

[0018] The edge server has strong task processing capabilities and caching capabilities, and can process calculation tasks and cache calculation results;

[0019] Cloud servers have a huge data storage capacity and computing power. As time accumulates, the data in the blockchain becomes increasingly large. The system diverts and stores the data in the blockchain on cloud servers to ensure sufficient storage space.

[0020] The edge computing task scheduling method and system in the industrial Internet of the present invention have the following advantages:

[0021] 1. A joint optimization model of task cache, blockchain, and task scheduling is established. In view of the characteristics of a large amount of computing tasks and high task similarity in the industrial Internet, a reliable cache mechanism for the industrial Internet assisted by blockchain is proposed, and the Improved Least Frequently Used (ILFU) algorithm is used to improve the content hit rate during data scheduling.

[0022] 2. The blockchain is combined with the edge computing architecture to address challenges such as insecure data sharing and data leakage caused by distributed edge servers. Through a unique data security mechanism, the blockchain can provide reliable cached data for the cache pool of edge servers, while ensuring more secure information interaction among distributed edge servers.

[0023] 3. Considering the influence of task cache and blockchain on the task scheduling location, the proposed Improved Genetic Algorithm-based Task Scheduling Policy Algorithm (IGA-TSPA) can effectively select a suitable task scheduling location according to delay constraints and resource constraints. Compared with traditional edge computing scheduling algorithms, the IGA-TSPA algorithm can obtain better scheduling effects in complex data environments. Description of the Drawings

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0025] Figure 1 It is a flowchart of the edge computing task scheduling method in the industrial Internet provided by the embodiment of the present invention;

[0026] Figure 2 It is a schematic diagram of a three-layer architecture combining blockchain and cloud-edge-end provided by the embodiment of the present invention.

[0027] Figure 3 Schematic diagram of cache update provided by an embodiment of the present invention;

[0028] Figure 4 Schematic diagram of the impact of the presence or absence of cache and blockchain on system cost provided by an embodiment of the present invention;

[0029] Figure 5 Schematic diagram of cost convergence comparison of different cache policies provided by an embodiment of the present invention;

[0030] Figure 6 Schematic diagram of cost convergence comparison of different solution algorithms provided by an embodiment of the present invention;

[0031] Figure 7 Schematic diagram of running time comparison of different solution algorithms provided by an embodiment of the present invention. Detailed implementation manners

[0032] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0033] Figure 1 Flowchart of the edge computing task scheduling method in the industrial Internet provided by an embodiment of the present invention. An embodiment of the present invention provides an edge computing task scheduling method in the industrial Internet, including:

[0034] S100, obtain a computing task.

[0035] Exemplarily, the computing task is a sequence, which includes a plurality of tasks generated sequentially.

[0036] S110, search in the cache of the edge server to see if there is a computing result that matches the computing task. If so, the edge server returns the computing result.

[0037] Exemplarily, when the computing task is generated, the policy server retrieves the data in the cache of the edge server by checking the task index r

[0037] ,

[0036] ,

[0035] , select ,

[0034] ,

[0038] , , dj , , , as required, and sets the retrieval time to T select (n), where n is the data size. When the retrieved data exists in the cache, the edge server directly returns the computing result. Since the amount of data as the computing result can be ignored compared with the data of the incoming computing task, the return time is ignored. At this time, the time for processing the computing task (uploading data time and retrieval time) is expressed as:

[0038]

[0039] Among them, represents the upload time of the computing task on the local device i and the edge server j. At this time, the consumption of the local device is:

[0040]

[0041] Among them, is the power when the local device is idle, is the energy consumption for uploading task data.

[0042] S120. If there is no matching calculation result in the cache of the edge server, search in the blockchain to see if there is a calculation result matching the computing task.

[0043] S130. If there is a calculation result in the blockchain, select the scheduling location according to the minimum cost; if not, select the scheduling location according to the delay and resources, and at the same time update the calculation result in the edge cache pool and the blockchain cache to obtain a reward.

[0044] Exemplarily, because the resources in the edge server are limited, the space in the cache cannot contain all task types. The system can use blockchain technology as an extension of the cache. When the required calculation result does not exist in the edge server, the policy server can obtain the required calculation result by searching the data in the blockchain, and the preset retrieval time is T find . Therefore, the total task processing time at this time is:

[0045]

[0046] At this time, the consumption of the local device is expressed as:

[0047]

[0048] When the delay is satisfied, the policy server does not retrieve the required calculation result in the caches of the blockchain and the edge server. At this time, the task must be processed on the edge server, and the number of CPU cycles processed per second by the edge server is f i e . At this time, the processing time of the task on the edge server is:

[0049]

[0050] The total processing time T 3 is expressed as:

[0051]

[0052] At this time, the consumption of the local device is:

[0053]

[0054] The selection of the above scheduling strategy must be completed within the deadline set for the task. If it exceeds the deadline, it may be discarded.

[0055] When there is no calculation result in the blockchain, the edge server processes the calculation task to obtain the corresponding real-time calculation result; stores the real-time calculation result in the edge cache pool and the blockchain cache, and updates the reward.

[0056] When storing the real-time calculation result in the cache of the edge server, it includes: S131, obtaining the real-time calculation result; S132, if the remaining space in the cache of the edge server is larger than the size of the real-time calculation result, then store the real-time calculation result in the cache of the edge server; S133, if the remaining space in the cache of the edge server is smaller than the size of the real-time calculation result, remove the calculation result with the lowest hit count from the calculation results stored in the cache of the edge server, and then store the real-time calculation result in the cache of the edge server.

[0057] During the process of task scheduling, whether the cache space is sufficient will affect the selection of the scheduling strategy. The hit rate of the calculation result will also affect the processing rate of the entire calculation task. Therefore, the present invention proposes an ILFU algorithm for multi-repetitive tasks in the industrial Internet to increase the hit rate of the cache.

[0058] The Least Frequently Used (LFU) algorithm selects the data with the lowest historical access frequency for elimination. The core idea of the LFU algorithm is "if the data has been accessed many times in the past, then it will be accessed more frequently in the future". However, the LFU algorithm also has some disadvantages. The LFU algorithm is powerless against sudden sparse traffic because the records that have been frequently accessed before have occupied the cache, and it is unlikely that unexpected traffic will be retained. At the same time, the LFU algorithm stores a large number of past accessed records, which occupies a large amount of space and may not be used in the future. This situation usually leads to a low cache utilization rate.

[0059] In view of the problems existing in the LFU cache policy, the present invention establishes a time interval task frequency table to record the hit times of calculation results in the cache within a period of time, and sets a hit rate threshold. The time interval task frequency table records the calculation task requirements for a period of time, calculates the hit rate of the calculation results in the cache during this period. If the hit rate is greater than this threshold, it means that the calculation results in the cache conform to the characteristics of the entire task. If the hit rate is less than this threshold, it means that the calculation results cached during this period cannot play the role of the cache well. The policy server will update the data in the cache. Specifically, multiple data with higher task frequencies can be selected from the time interval task frequency table and filled into the cache to ensure that the content hit rate in the next stage maintains a normal cache effect.

[0060] When the computing task R ∈ {R1, R2, R3... R n} is scheduled to the edge server, the policy server searches for the calculation results based on the requirements of the computing task. Each cache task is represented by y i ={r dj , o i , c i}. Here, r dj represents the cache index and also the task index, o i represents the number of cache hits, and c i represents the space occupied by the task, which is convenient for replacing cache tasks in the future. To handle multiple different types of tasks and ensure the data security of the cache, the present invention requires the cache to cache different types of data as much as possible. At the same time, the data source of the cache can only be the calculation results processed by this edge server and the calculation results in the blockchain. The cache source is as shown in Figure 3 .

[0061] The present invention defines y i ∈ {0, 1} to represent whether the calculation result corresponding to the computing task R i exists in the cache. When performing cache operations, the cache capacity is limited. The total capacity limit of cache tasks is:

[0062]

[0063] Among them, C represents the maximum capacity of the edge server cache pool. When the edge server finishes processing the computing task and obtains the real-time calculation result, if the cache of the edge server has sufficient cache space, that is, the remaining cache space of the edge server is greater than the size of the real-time calculation result, the calculation result can be directly cached at this time. The size of the task data that can be cached is limited to:

[0064]

[0065] When the cache space is insufficient, that is, the remaining cache space of the edge server is less than the size of the real-time calculation result, the policy server will find the calculation result with the smallest o i value in the cache, that is, the calculation result with the lowest hit frequency during this period, and replace it. Remove this calculation result from the cache, and MINo i represents the space occupied by the calculation result with the smallest number of hits. The data size of the calculation result that can be cached is expressed as:

[0066]

[0067] In the present invention, by establishing a time interval task frequency table, the hit result situation of the calculation results within t time intervals is recorded and the initial threshold Ω is set. Assume that the hit rate in the time period [t - 2n, t - n] is ψ. The policy server will determine the hit rate situation and the predetermined hit rate threshold at time t. If ψ < Ω, it means that the distribution of the calculation results stored in the cache is unreasonable and the content of the cache needs to be updated again. The policy server will update and import several calculation results with higher frequencies in the time interval task frequency table from the blockchain into the cache to generate new cache data.

[0068] S140. Based on the edge cache pool and the blockchain cache, establish a scheduling model according to the information of the calculation task.

[0069] Exemplarily, represent the calculation task as R ij ={r ij ,r wj ,r dj ,t max},R ij represents the calculation task generated by the local device i at time j, r ij represents the data size of the input of the calculation task, such as code or required resources, r wj represents the computing power requirement of the calculation task, r dj represents the task index required by the calculation task, t max represents the maximum latency tolerated by the calculation task.

[0070] When establishing the scheduling model in S140, the following method can be adopted: S141. Obtain the task index in the calculation task; S142. Determine whether the calculation result exists in the cache of the edge server or the blockchain according to the task index; S143. Establish a scheduling model according to the determined location of the calculation result.

[0071] Due to the limitations of the energy and computing power of the local device, the system will, under the maximum latency allowed for the computing task, selectively schedule the computing task to the edge server for processing according to the cost issue. The present invention defines the task scheduling strategy as x i,j ∈{0, 1}, where x i,j = 0 represents that the computing task will be computed on the local device, and x i,j = 1 represents that the computing task i will be processed on the edge server j.

[0072] When the computing task is scheduled to the edge server for processing, the edge server needs to have sufficient resources. After the computing task is processed, the edge server will return the computing result. Since the computing result is very small, the result return time is ignored here. Assume that P ij represents the uplink transmission capacity from device i to edge server j, and the uplink rate V ij :

[0073]

[0074] where σ 2 represents the noise power, |r i | -a represents the path fading (which changes with the path length and causes obvious attenuation), and |h i | 2 represents the mutual influence between signals. Set According to the attributes of the computing task, the upload time of the link (data size / uplink rate) can be obtained as:

[0075]

[0076] During the upload process of the computing task, the energy consumption of the local device is:

[0077]

[0078] where ζ is the efficiency of the device's transmitting power amplifier. Let represent the local device consumption: is the energy required for the task to run on the local device, is the energy consumed by the local device for transmission, is the consumption when the local device is idle.

[0079] When the computing task is processed on the local device, the computing power of the local device is f i l , and during the processing, the consumption of the computing task on the local device i is:

[0080] E0 = k(f il ) 2 r w,j (14)

[0081] where k = 10 -26 is a coefficient depending on the chip architecture, and at this time f i l will affect the calculation execution time and energy consumption.

[0082] When the computing task is processed on the edge server, assume that the data stored in the cache of the edge-side server is Y i =(y1, y2... y n ). Whether the calculation result required by the task exists in the cache will have different impacts on the system energy consumption. Next, different cache situations will be classified and described.

[0083] In the environment of the industrial Internet, a suitable scheduling strategy can be obtained by considering conditions such as the delay constraint of the computing task, the computing resources and storage resources of local devices, etc. According to the above discussion, the total energy consumption of the entire system can be expressed as:

[0084]

[0085] By obtaining the minimized cost consumption through the optimal offloading decision and combining the reward for updating the calculation result in the cache, the optimal system cost can be obtained:

[0086]

[0087]

[0088]

[0089]

[0090]

[0091]

[0092] ∑e i ≤E (f)

[0093] ∑f i ≤F (g)

[0094] (T off +T mec )≤T deadline (h)

[0095] z i y i (T find +T off) ≤ T deadline (i)

[0096] Among them, ε represents the conversion coefficient of energy consumption and cost, represents the reward for updating the calculation result of the edge server. Formulas (a), (b), and (c) represent the migration location of the calculation result, whether the calculation result exists in the cache, and whether the calculation result exists in the blockchain respectively. Formula (d) represents that the maximum cache capacity of the calculation result cannot exceed the cache capacity of the edge server. Formula (e) represents that the resource processing capacity required by the calculation task cannot exceed the total resources of the edge server. (f) and (g) represent that the computing power and power of the edge server are limited. (h) and (i) represent that the processing and search time of the calculation task cannot exceed the deadline of the predetermined task.

[0097] S150. Adopt the task scheduling strategy algorithm based on the improved genetic algorithm to solve the scheduling model and determine whether the calculation task is scheduled to the edge server.

[0098] Exemplarily, in the process of selecting the solution algorithm, it is found that this is an NP-hard problem. Given the size v of n products i and the profit p i , where i ∈ N, set the maximum capacity of the knapsack to V. The goal is to find a set s ∈ N to maximize the profit of the knapsack, which can be expressed as:

[0099]

[0100] Analogize the scheduling problem to a knapsack problem. Assume there are N calculation tasks and the time to complete the tasks is T all . The calculation task can be executed on the local device or run on the edge server. Set t i , e i to be the time and energy consumption generated by migrating and executing the calculation task r i respectively. The optimization problem can be expressed as:

[0101]

[0102] Redefine Then the problem can be expressed as:

[0103]

[0104] Analogize T all to V, and w i to p i . Therefore, it can be said that the scheduling problem is also an NP-hard problem.

[0105] Edge computing task scheduling with caching is a multi-constraint optimization problem. As the number of tasks and cached data increases, the solution space will be extremely large and complex. To address this issue, the present invention proposes an improved genetic algorithm to find the optimal feasible solution in the vast solution space. The genetic algorithm is a global heuristic algorithm that follows the law of "survival of the fittest" in nature, selects feasible solutions, and finally obtains the optimal solution to the problem. However, the traditional genetic algorithm has a large solution space, a long iteration period, and a slow convergence speed, which cannot well meet the requirements of the industrial Internet. Therefore, the present invention improves the population initialization and mutation of the genetic algorithm to make the algorithm more adaptable to the model of the present invention, thereby improving the rate of the algorithm to find the optimal solution.

[0106] At the beginning of the genetic algorithm, it is necessary to initialize the population. The population consists of multiple chromosomes, and each chromosome represents a feasible solution to the task, that is, all scheduling strategies. A chromosome consists of multiple genes, and each group of chromosomes represents a feasible solution. In this system, a gene represents the offloading location of the task. The present invention uses 0-1 binary encoding, where 0 represents that the task is executed on the local device, and 1 represents that the task is processed on the edge server. When the IGA-TSPA algorithm performs the initialization operation, it first makes a judgment on the computing task. The IGA-TSPA algorithm will calculate the processing time of the computing task on the local device according to the number of cycles r wj required for the computing task to be processed and the processing frequency f i l of the local device. Next, it checks the resource requirements of the computing task. If it exceeds the set demand threshold, the policy server will set the initial value of the computing task to 1, indicating that this computing task may be processed faster at the edge node. The IGA-TSPA algorithm speeds up the search frequency of tasks by shrinking the solution space.

[0107] Fitness is the criterion for evaluating the quality of an individual. As the number of population iterations increases, individuals with larger fitness function values are more likely to survive. The selection of the fitness function of the present invention is determined by equation (16). The algorithm deletes individuals with excessive latency, excessive capacity, and excessive power consumption during the stage of fitness function selection by restricting data such as time, cache status, and power, and obtains feasible solutions that meet the conditions.

[0108] The selection operation is the process of selecting individuals from the previous generation population to the next generation population. Generally, individuals are selected according to the distribution of individual fitness. In the process of natural evolution, more adaptable species can obtain better resources, and thus can obtain the right to survive and reproduce, and such individual genes will be retained. The system will select individuals with higher fitness function values and use the roulette wheel method. The probability of an individual being selected is proportional to the size of its fitness function value.

[0109] The fitness f of an individual i i , with the population size GENE, the probability formula for i to be selected is:

[0110]

[0111] In the genetic evolution algorithm, mutation is a method that may skip local optimal solutions. The mutation operation can accelerate the convergence of the optimal solution, maintain population diversity at the same time, and prevent "premature maturity". However, mutation is an uncontrollable and random operation, so there are some cases of ineffective mutation in practice, which instead makes the value of the fitness function smaller. The operation of the present invention at the mutation point is as follows:

[0112] 1. Randomly select a computing task in the sequence of computing tasks as the mutation point. First, judge whether the selected computing task needs to mutate through the mutation probability Q y

[0113] 2. If it is determined that mutation is required, then check the resource requirement of the computing task at the mutation point, judge whether it exceeds the limit of the mutation threshold α2, and decide whether to finally complete the mutation.

[0114] The embodiment of the present invention also provides an edge computing task scheduling system in the industrial Internet, including:

[0115] A policy server, configured to obtain computing tasks, establish a scheduling model according to the information of the computing tasks, solve the scheduling model, and determine whether the computing tasks are scheduled to the edge server;

[0116] When the computing tasks are scheduled to the edge server, the policy server searches in the cache of the edge server to see if there is a computing result matching the computing tasks. If there is a computing result in the cache of the edge server, the edge server returns the computing result; if there is no computing result in the cache of the edge server, the policy server searches in the blockchain to see if there is a computing result matching the computing tasks; if there is a computing result in the blockchain, the blockchain returns the computing result.

[0117] Such as Figure 2 ​As shown in the figure, in the industrial Internet scenario, the entire system is divided into three layers - the local device layer, the edge server layer, and the cloud layer. The policy server is located in the edge server layer. The local devices in the local device layer are composed of various factory devices, with a small amount of computing power, data collection ability, and partial information sorting ability. The devices in the local device layer are represented as N = {1, 2, 3... n}, and computing tasks are generated in the local device layer. The edge server layer is composed of various edge servers, which can be represented as M = {1, 2, 3... m}. The edge servers have stronger task processing capabilities and caching capabilities than local devices. At this time, the edge servers can not only process computing tasks, but also be responsible for updating the newly generated cached computing results to the blockchain to ensure that the cache can match more task types. The third layer is the cloud layer, which has a huge data storage capacity and computing power. Considering that as time accumulates, the data in the blockchain becomes larger and larger, the policy server can split and store the data in the blockchain in the cloud to ensure sufficient storage space.

[0118] Experimental results show that

[0119] Next, the performance of the proposed model and algorithm will be evaluated. The specific experimental parameters are shown in Table 1.

[0120] Table 1 Parameter Settings

[0121]

[0122]

[0123] According to the actual situation, the computing capabilities of local devices, edge servers, and the cloud are set to 1 GHz, 4 GHz, and 10 GHz respectively. The cloud server has the strongest computing power, followed by the edge server, and the local device has the weakest computing power. At the same time, it is stipulated that the power of the local device is 0.3 W during idle time, 0.9 W when the local device is executing tasks, and 1.2 W during the transmission stage. In addition, [100, 300] computing tasks with a task volume of [about 20, 40] Mbit are randomly selected. The maximum allowable delay for each computing task is within [1, 5] s. At the same time, the channel is specified, the channel noise is -100 dpm, and the maximum transmission power is 20 dpm. To improve the experimental results, the present invention has conducted a large number of simulation experiments to obtain results that meet expectations.

[0124] To prove the effectiveness of the present invention, as Figure 4In (a), the present invention compares the cost convergence of computing tasks with and without caching. The upper and lower curves respectively represent the cases without caching and with caching. In the case of caching, both the convergence speed and the convergence of the algorithm are better than those without caching. Without considering the return consumption and the relatively small cache search time, the issues affecting the system cost are only the task upload link cost and the edge computing cost. The existence of caching reduces the edge computing cost to a certain extent, especially in industrial scenarios, and the effect of cost control is more significant.

[0125] The size of the cache space determines whether the cache types are complete. The existence of the blockchain is equivalent to backing up all processing results. Therefore, when the blockchain exists, the system cost will also decrease. The present invention compares Figure 4 (a) and (b). Without considering the retrieval cost, since the blockchain has a larger capacity than the cache of the edge server and can match more task types, in Figure 4 (b), the upper and lower curves respectively represent the cases without blockchain caching and with blockchain caching. Therefore, introducing the blockchain and caching in the edge computing architecture is of great significance in meeting the latency requirements and reducing costs, and can meet the latency requirements of the industrial Internet.

[0126] The present invention compares four cache replacement algorithms, namely the First Input First Output (FIFO) algorithm, the Least Recently Used (LRU) algorithm, the LFU algorithm, and the random algorithm. As Figure 5 shown, with a fixed cache size, it can be seen that compared with the other four algorithms, the cost after convergence of the ILFU algorithm is smaller under the same number of iterations. Among them, Figure 5 the five curves from top to bottom in are the random algorithm, the LFU algorithm, the FIFO algorithm, the LRU algorithm, and the ILFU algorithm proposed by the present invention.

[0127] On the basis of determining the cache policy, the present invention analyzes the performance of IGA-TSPA. As Figure 6 shown, by comparing the IGA-TSPA algorithm with several common global search algorithms such as the traditional genetic algorithm, the simulated annealing algorithm, and the hill climbing algorithm, the IGA-TSPA algorithm can obtain good convergence results. Figure 6The five curves from top to bottom are the Hillclimbing Algorithm (HA), Simulated Annealing (SA), Random Algorithm (RA), Genetic Algorithm (GA), and the IGA-TSPA algorithm proposed by the present invention. The cost of the IGA-TSPA algorithm is approximately 10% lower than that of the traditional genetic algorithm. Since the genetic algorithm randomly generates a large number of solutions, the feasible solution space is huge, making it difficult to search for the optimal solution. When initializing the population, the IGA-TSPA restricts the initialization of some tasks, reducing the time for the algorithm to select the best. At the same time, during the mutation process, the IGA-TSPA algorithm avoids some useless mutations, thereby improving the convergence level of the algorithm by restricting the mutation operation. As Figure 7 shown, the present invention compares the running time of the algorithms. From the results, the time of the random algorithm is the lowest, and the algorithm times of the GA algorithm and the IGA-TSPA algorithm are moderate. Because the IGA-TSPA algorithm reduces the time for selecting the best in the initialization of the population, the algorithm reaches the optimal value faster, proving that the algorithm improvement is effective.

[0128] To solve the task scheduling problem caused by large amounts of data, complex task types, and strong device heterogeneity in the modern industrial environment, the present invention establishes a joint optimization model of task caching, blockchain, and task scheduling, and uses blockchain technology to propose a three-layer architecture combining industrial devices, edge computing, and cloud computing to ensure the security of information transmission and data storage. The combination of blockchain and the caching mechanism greatly improves the data security of the industrial Internet and effectively reduces latency and costs. In addition, the present invention proposes the ILFU algorithm to improve the content hit rate of the edge computing cache pool and further reduce task latency. Finally, on the basis of ensuring the security of caching and data sharing, considering the influence of task caching, blockchain, and blockchain rewards on the scheduling strategy, IGA-TSPA is proposed to optimize the task processing cost problem. Under the condition of ensuring the time delay, IGA-TSPA can obtain a scheduling strategy that meets the situation at the lowest cost. A large number of simulation experiments prove the effectiveness of the model and the algorithm.

[0129] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the present invention.

[0130] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these modifications and variations.

Claims

1. An edge computing task scheduling method in the industrial Internet, characterized in that Including: Obtain a computing task; Search in the cache of the edge server to check if there is a computing result that matches the computing task. If there is, the edge server returns the computing result; If there is no matching computing result in the cache of the edge server, search in the blockchain cache to check if there is a computing result that matches the computing task; If the computing result exists in the blockchain cache, select a scheduling location based on the minimum cost; if not, select a scheduling location based on latency and resources, and at the same time update the computing result in the edge cache pool and the blockchain cache to obtain a reward; Based on the edge cache pool and the blockchain cache, establish a scheduling model according to the information of the computing task; Use a task scheduling strategy algorithm based on an improved genetic algorithm to solve the scheduling model and determine whether the computing task is scheduled to the edge server; The establishing a scheduling model according to the information of the computing task based on the edge cache pool and the blockchain cache includes: Obtain the task index in the computing task; Determine whether the computing result exists in the cache of the edge server or the blockchain cache according to the task index; Establish the scheduling model according to the determined location where the computing result is located; When using the task scheduling strategy algorithm based on the improved genetic algorithm to solve the scheduling model, it is necessary to initialize the population. During initialization, compare the resource requirement of the computing task with a set requirement threshold. If the resource requirement exceeds the requirement threshold, determine to schedule the computing task to the edge server.

2. The edge computing task scheduling method in the industrial Internet according to claim 1, wherein When the computing result does not exist in the blockchain cache, the edge server processes the computing task to obtain a corresponding real-time computing result; Store the real-time computing result in the edge cache pool and the blockchain cache, and update the reward.

3. The edge computing task scheduling method in the industrial Internet according to claim 2, wherein The storing the real-time computing result in the edge cache pool and the blockchain cache includes: Obtain the real-time computing result; If the remaining space in the cache of the edge server is greater than the size of the real-time computing result, store the real-time computing result in the cache of the edge server; If the remaining space in the cache of the edge server is less than the size of the real-time computing result, remove the computing result with the lowest hit count from the cache stored in the edge server, and then store the real-time computing result in the cache of the edge server.

4. The edge computing task scheduling method in the industrial Internet according to claim 2, wherein It also includes: Establish a time interval task frequency table, which records the hit rate of the computing results stored in the cache of the edge server within a period of time; Compare the hit rate recorded in the time interval task frequency table with a set hit rate threshold. If the hit rate recorded in the time interval task frequency table is less than the hit rate threshold, update the computing results stored in the cache of the edge server.

5. The edge computing task scheduling method in the industrial Internet according to claim 1, characterized in that When using the task scheduling strategy algorithm based on the improved genetic algorithm to solve the scheduling model, the mutation operation on an individual includes: Determine whether the individual needs to mutate according to the mutation probability; When it is determined that the individual needs to mutate, query the individual's demand for the resource quantity; If the individual's demand for the resource quantity exceeds the set mutation threshold, complete the mutation operation on the individual.

6. A system for applying the edge computing task scheduling method in the industrial Internet according to any one of claims 1-5, characterized in that, Including: The policy server is used to obtain a computing task and search in the cache of the edge server to see if there is a computing result that matches the computing task. If there is, the edge server returns the computing result. If there is no matching computing result in the cache of the edge server, the policy server searches in the blockchain cache to see if there is a computing result that matches the computing task; If the computing result exists in the blockchain cache, the policy server selects a scheduling location according to the minimum cost; If not, the policy server selects a scheduling location according to the latency and resources, and at the same time updates the computing result in the edge cache pool and the blockchain cache to obtain a reward; Based on the edge cache pool and the blockchain cache, the policy server establishes a scheduling model according to the information of the computing task, and then uses a task scheduling strategy algorithm based on an improved genetic algorithm to solve the scheduling model to determine whether the computing task is scheduled to the edge server.

Images