An edge-side collaborative method for time series data in 5g private network environment

Abstract

The application belongs to the field of edge computing, and particularly relates to a terminal network extension device and an edge device cooperative caching method, mainly applied to reasonable allocation and utilization of resources in edge caching, and the method comprises the following steps: constructing a 5G private network, including a wireless fusion network shunt, an edge computing server MEC, a WiFi access point, a 5G CPE access point, and terminal devices accessing the 5G private network through the 5G CPE; deploying an Agent on the 5G CPE in the 5G private network to collect time sequence data information requested by each intelligent terminal device, and sending the time sequence data information to the MEC after evaluating the content preference; the MEC solves a minimum cost end-edge cooperative caching strategy under the 5G private network environment according to a constructed edge caching cost model and the obtained content preference; and the MEC requests corresponding content from the core network according to the cooperative caching strategy and allocates the corresponding content to different 5G CPEs, so that the caching scheme is updated according to the request change of the intelligent terminal device.

Description

Technical Field

[0001] This invention belongs to the field of edge computing, and specifically relates to a collaborative caching method for terminal network extension devices and edge devices, which is mainly applied to the rational allocation and utilization of resources in edge caching. Background Technology

[0002] With the application and popularization of 5G communication technology, smart terminal devices in industrial IoT scenarios are typically connected to the core network via routers, 5G CPEs, and other similar devices. Consequently, the industrial IoT has driven a massive influx of related industrial time-series data into the network. Unlike data on traditional cloud servers, this data is mostly updated in real time as tasks are executed, impacting network load and request latency. Therefore, reducing network resource consumption and improving network performance has received widespread attention and importance in recent years.

[0003] Edge caching technology, as an important research direction in edge computing, shows promise in reducing data request latency and improving network performance. By pre-caching core network data in 5G CPEs closer to terminal devices, the data transmission request distance of devices is greatly shortened, thereby effectively reducing data request latency. Furthermore, it effectively avoids duplicate data requests and responses being transmitted on the core network, reducing network load.

[0004] However, the resources on both terminal devices, such as network extension devices closer to the user, and edge devices, such as edge servers bypassing base stations, are relatively limited and cannot cache all the data needed by the user. Therefore, determining the most profitable content to cache in the current environment and designing a reasonable edge caching strategy to make efficient use of the limited resources on edge devices and minimize the cost of the caching process is a problem that researchers urgently need to solve. Summary of the Invention

[0005] An edge-to-edge collaborative caching method for time-series data in a 5G private network environment is characterized by the following steps:

[0006] S1. Construct a 5G private network, including a wireless converged network splitter, an edge computing server MEC, a WiFi access point, and a 5G CPE access point. Terminal devices access the 5G private network through the 5G CPE.

[0007] S2. Deploy an Agent on the 5G CPE within the 5G private network to collect time-series data information requested by each smart terminal device, evaluate its content preferences, and then send it to the MEC.

[0008] S3 and MEC solve for the minimum cost edge-to-edge collaborative caching strategy in the 5G private network environment based on the constructed edge caching cost model and the obtained content preferences.

[0009] S4. MEC requests the corresponding content from the core network according to the collaborative caching strategy in step S3 and allocates it to different 5GCPEs, thereby achieving the purpose of updating the caching scheme according to the changes in requests from smart terminal devices.

[0010] Furthermore, the 5G private network in step S1 specifically includes:

[0011] S11. Deploy a wireless converged network splitter on the 5G base station side and connect it to the local network through the wireless converged network splitter. Deploy MEC in bypass mode of the converged network splitter.

[0012] S12. All terminals of a 5G private network are connected to the network by combining multiple 5G CPEs. The 5G CPEs and MECs communicate with each other through the wireless converged network splitter to build an end-edge collaborative network.

[0013] S13. Utilizing certain storage resources in 5G CPE and MEC, smart terminal devices can pre-cache requested file data based on the principle of proximity.

[0014] Furthermore, a collaborative caching method for time-series data in a 5G private network environment is characterized in that step S2 specifically includes:

[0015] The Agent deployed on the S21 and 5G CPE is an executable program running under the embedded system. It continuously collects information such as the location, time, and frequency of requests for time-series data from terminal devices connected to the current 5G CPE.

[0016] S22. Perform preference evaluation on the collected time-series data and report the relevant information to MEC.

[0017] Furthermore, the time-series data in step S21 refers to files that are continuously updated over time. The time-series data may be requested by a large number of terminal devices in the short period of time after it is generated, but it will not be requested by terminal devices after a certain interval. This interval is called the maximum tolerance time of the smart terminal device for the time-series data.

[0018] Furthermore, the preference evaluation of time-series data in step S22 includes:

[0019] S221, using This indicates that all smart terminal devices access time-series data. The total number of requests is then Global content popularity It can be represented as:

[0020] ;

[0021] Whenever a smart terminal device After initiating the request, then... Update;

[0022] S222, using Record time series data The most recent update time is typically only for time-series data on each smart terminal device. Make at most one request, using Indicates intelligent terminal device Have you requested time-series data? If a request has already been made, then ,otherwise ;

[0023] S223, Use This indicates that smart terminal devices access time-series data. The average maximum tolerance time is in the time slot. Smart terminal devices For time series data preferences It can be represented as:

[0024] ;

[0025] Furthermore, step S3 specifically includes:

[0026] S31. Construct an edge caching cost model;

[0027] S32. MEC uses the time-series data content preferences obtained in step S22 to solve for the optimal caching scheme in the edge-end collaborative caching network.

[0028] Furthermore, the cache cost model constructed in step S31 is as follows:

[0029] ;

[0030] In the formula, This represents the average cost of all requests within the local network. This represents the cost for a terminal device to retrieve data from its local 5G CPE once. This represents the cost for a terminal device to retrieve data from the MEC once. This represents the cost for a terminal device to obtain data from another 5G CPE once; This represents the cost for a terminal device to obtain data from the core network once. This indicates the number of times the terminal device obtains data from the local 5G CPE; This indicates the number of times the terminal device obtains data from the MEC; This indicates the number of times the terminal device obtains data from other 5G CPEs; This indicates the number of times the terminal device obtains data from the core network.

[0031] Furthermore, step S32, based on the improved fireworks algorithm, obtains the optimal end-edge collaborative caching strategy. The specific steps are as follows:

[0032] S321. Initialize using a greedy algorithm. Using different caching strategies as a fireworks population, while ensuring sufficient caching resources, priority is given to caching time-series data that has a higher preference and has not yet been cached in the edge-to-edge collaborative network.

[0033] S322. Iterate through each firework and generate explosive sparks, and perform targeted mutations on the firework population according to a certain probability;

[0034] S323. Calculate the adaptability of different fireworks based on the cost model in step S31. The lower the cost of the fireworks, the better their adaptability.

[0035] S324, Select the one with the highest fitness. A firework entered a new population;

[0036] S325. If the maximum number of iterations is not met, proceed to step S322; otherwise, obtain the firework individual with the highest fitness as the optimal solution.

[0037] Furthermore, step S322, the directed mutation of the fireworks population, includes randomly selecting the row to be mutated. , will be done All existing values ​​are set to 0, select the route. The most preferred content is cached, all the way up to the next line. Insufficient cache resources.

[0038] Furthermore, the specific process by which MEC distributes requested content to the corresponding 5G CPE according to the collaborative caching strategy includes:

[0039] S41. MEC sequentially searches for the current time-series data to be cached in the collaborative cache set in the edge-edge collaborative network using a HASH algorithm. If no match is found, it jumps to step S42.

[0040] S42 and MEC initiate a content request to the core network based on the requested time-series data set;

[0041] S42, MEC allocates the requested content to the corresponding 5G CPE according to the collaborative caching strategy obtained in step S32. Attached Figure Description

[0042] Figure 1A preferred embodiment of the edge-end collaborative system architecture diagram is provided for this invention;

[0043] Figure 2 This is a flowchart illustrating the caching strategy generated by the improved fireworks algorithm of this invention.

[0044] Figure 3 This is a diagram illustrating the process of data acquisition by the intelligent terminal device of the present invention. Detailed Implementation

[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] The network architecture of this invention is as follows: Figure 1 As shown, the specific functions of each part in this architecture are detailed below:

[0047] (1) The local network includes various types of smart terminal devices, 5G CPEs, etc. Here, 5G CPE can be a CPE device in a 5G private network environment, and smart terminal devices access the network through the CPE device. The CPE device can communicate with the base station, and the requests received by the base station are forwarded to the core network or MEC by the UPF.

[0048] (2) In the application scenario of the present invention, the MEC can communicate with the CPE device through the UPF and control the CPE device in reverse, thereby realizing the collaboration between each CPE device and the MEC.

[0049] (3) In the application scenario of this invention, the request initiated by the smart terminal device is first processed by the local CPE device connected to it. The request packets for accessing the MEC and accessing the core network are usually access requests sent by the smart terminal device through the CPE device. After receiving the request from the CPE device, UPF will forward it to the MEC, other CPE devices or the core network according to the different target addresses of the request.

[0050] MEC is responsible for allocating its own cache resources and the cache resources on all CPE devices. Upon receiving request information from all CPE devices regarding terminal devices, MEC processes the information, evaluates the preference for time-series data, and then uses an improved fireworks algorithm to formulate caching schemes for different CPE devices. The workflow of the improved fireworks algorithm is as follows: Figure 2 As shown, starting from step 101, the specific steps are as follows:

[0051] In step 102, N fireworks are initialized as the initial fireworks population. Each firework is represented by a multi-dimensional vector, where each row of the vector represents a caching scheme on a CPE device. Randomly initializing the population as in the original algorithm would result in generally low fitness of the initial population, affecting the iteration effect. This invention, considering the characteristics of collaborative caching, incorporates a greedy approach while ensuring that the same content is cached as little as possible. When formulating a caching scheme for any CPE device, it prioritizes caching data with higher preference, and to ensure efficient use of cache resources, the same data is cached only once on the local network whenever possible.

[0052] In step 103, the fitness of each firework is calculated. Each firework represents a caching scheme for the entire local network. By combining the preferences of time-series data, the caching cost of the caching scheme can be calculated. The lower the caching cost of the firework, the better its fitness, and the more explosion sparks it can produce.

[0053] In step 104, the number of explosive sparks and the explosion amplitude that each firework can produce are calculated based on the adaptability of the fireworks, thereby generating a certain number of explosive sparks;

[0054] In step 105, this invention incorporates a spark-directed mutation step. First, rows to be mutated are randomly selected based on the fitness of the spark. Then, without exceeding cache resource limits, the time-series data with the highest preference is cached.

[0055] In step 106, based on the fitness of the fireworks, the top N fireworks with the highest fitness are selected as the new population;

[0056] In step 107, the optimization of the fireworks algorithm is judged. If the number of iterations reaches the threshold, step 108 is executed; otherwise, step 103 is executed.

[0057] In step 108, the firework with the best fitness is selected as the caching scheme for all caching devices in the local network, and then the process proceeds to step 109 to end.

[0058] This embodiment processes content requests from terminal devices based on the principle of proximity. Figure 3 The process of a terminal device hitting content is described, starting from step 201 where the terminal device initiates a request to the local CPE device. The specific process is as follows:

[0059] In step 202, the CPE device first retrieves content from the local machine. If the target is not found, step 203 is executed; otherwise, step 212 is executed.

[0060] In step 203, the CPE device forwards a request to the MEC connected to the base station, and then proceeds to step 204;

[0061] In step 204, the MEC on the base station side retrieves content locally. If the target is not found, step 205 is executed; otherwise, step 211 is executed.

[0062] In step 205, the base station initiates a content retrieval request to other CPE devices, and then proceeds to step 206;

[0063] In step 206, other CPE devices perform content retrieval. If content is found, step 207 is executed.

[0064] In step 207, other CPE devices return the search results to MEC. If the content is found, step 211 is executed; otherwise, step 208 is executed.

[0065] In step 208, the MEC forwards the request to the core network, and then proceeds to step 209;

[0066] In step 209, the core network retrieves content and proceeds to step 210;

[0067] In step 210, the core network returns the retrieved content to the MEC, and then proceeds to step 211;

[0068] In step 211, the MEC returns the content to the local CPE device;

[0069] In step 212, the local CPE device returns the content to the terminal device, and the request ends.

[0070] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

A 1.5G private network environment provides an edge-to-edge collaborative caching method for time-series data, characterized in that... Specifically, the following steps are included: S1. Construct a 5G private network, including a wireless converged network splitter, an edge computing server (MEC), a WiFi access point, and a 5G CPE access point. Terminal devices access the 5G private network through the 5G CPE. S2. Deploy an Agent on the 5G CPE within the 5G private network to collect time-series data information requested by each smart terminal device, evaluate its content preferences, and then send it to the MEC. S3 and MEC solve for the minimum cost edge-to-edge collaborative caching strategy in the 5G private network environment based on the constructed edge caching cost model and the obtained content preferences. S4. MEC requests the corresponding content from the core network according to the collaborative caching strategy in step S3 and allocates it to different 5G CPEs, thereby achieving the purpose of updating the caching scheme according to the changes in requests from smart terminal devices. Step S2 specifically includes: The Agent deployed on the S21 and 5G CPE is an executable program running under the embedded system. It continuously collects information on the location, time, and frequency of requests for timing data from terminal devices connected to the current 5G CPE. S22. Evaluate the preferences of the collected time-series data and report the relevant information to MEC; The time-series data in step S21 refers to files that are constantly updated over time. The time-series data may be requested by a large number of terminal devices in a short period of time after it is generated, but it will not be requested by terminal devices after a certain interval. This interval is called the maximum tolerance time of smart terminal devices for time-series data. The preference evaluation of time series data in step S22 includes: S221, using q r Let represent the total number of requests for time-series data r from all smart terminal devices. Then, the global content popularity cp of r is... r Represented as: ； Whenever a smart terminal device initiates a request to r, it then requests q. r Update; S222, using tb r Record the most recent update time of the time-series data r. Each smart terminal device typically makes at most one request for the time-series data r. This indicates whether the smart terminal device u has requested the time-series data r. If it has, then... ,otherwise ; S223, Use TA r Let represent the average maximum tolerance time of a smart terminal device for time-series data r. Then, in time slot t, the preference of smart terminal device u for time-series data r is... The preference is expressed as: 。 2. The edge-end collaborative caching method for time-series data in a 5G private network environment according to claim 1, characterized in that, Step S1 specifically includes: S11. Deploy a wireless converged network splitter on the 5G base station side and connect it to the local network through the wireless converged network splitter. Deploy MEC in bypass mode of the converged network splitter. S12. All terminals of a 5G private network are connected to the network by combining multiple 5G CPEs. The 5G CPEs and MECs communicate with each other through the wireless converged network splitter to build an end-edge collaborative network. S13. Utilizing certain storage resources in 5G CPE and MEC, smart terminal devices pre-cachate requested file data based on the principle of proximity.

3. The edge-to-edge collaborative caching method for time-series data in a 5G private network environment according to claim 1, characterized in that, Step S3 specifically includes: S31. Construct an edge caching cost model; S32. MEC uses the time-series data content preferences obtained in step S22 to solve for the optimal caching scheme in the edge-end collaborative caching network.

4. The edge-to-edge collaborative caching method for time-series data in a 5G private network environment according to claim 3, characterized in that... The cache cost model constructed in step S31 is as follows: ； In the formula, This represents the average cost of all requests within the local network. This represents the cost for a terminal device to retrieve data from its local 5G CPE once. This represents the cost for a terminal device to retrieve data from the MEC once. This indicates the cost for a terminal device to obtain data from another 5GCPE once; This represents the cost for a terminal device to obtain data from the core network once. This indicates the number of times the terminal device obtains data from the local 5G CPE; This indicates the number of times the terminal device obtains data from the MEC; This indicates the number of times the terminal device obtains data from other 5G CPEs; This indicates the number of times the terminal device obtains data from the core network.

5. The edge-end collaborative caching method for time-series data in a 5G private network environment according to claim 3, characterized in that, Step S32 uses the improved fireworks algorithm to solve for the optimal edge-to-edge collaborative caching strategy. The specific steps are as follows: S321. Initialize using a greedy algorithm. Different caching strategies are used as fireworks populations. Under the premise of ensuring sufficient caching resources, priority is given to caching time-series data with higher preference that has not yet been cached in the edge-to-edge collaborative network. S322. Iterate through each firework and generate explosive sparks, and perform targeted mutations on the firework population according to a certain probability; S323. Calculate the adaptability of different fireworks based on the cost model in step S31. The lower the cost of the fireworks, the better their adaptability. S324, Select the one with the highest fitness. A firework entered a new population; S325. If the maximum number of iterations is not met, proceed to step S322; otherwise, obtain the firework individual with the highest fitness as the optimal solution.

6. The method for solving the optimal edge-to-edge collaborative caching strategy for time-series data in a 5G private network environment according to claim 5, characterized in that, Step S322, the directed mutation of the fireworks population, includes randomly selecting the row to be mutated. , will be done All existing values ​​are set to 0, then select a route. The most preferred content is cached, all the way up to the next line. Insufficient cache resources.

7. The edge-end collaborative caching method for time-series data in a 5G private network environment according to claim 5, characterized in that, In step S4, the MEC distributes the requested content to the corresponding 5G CPE according to the collaborative caching strategy. The specific process includes: S41. MEC sequentially searches for the current time-series data to be cached in the collaborative cache set in the edge-edge collaborative network using a HASH algorithm. If no match is found, it jumps to step S42. S42 and MEC initiate a content request to the core network based on the requested time-series data set; S43. MEC allocates the requested content to the corresponding 5G CPE according to the collaborative caching strategy obtained in step S32.

Images