A method for placing content in edge cache

Abstract

The present invention discloses a method for placing content in an edge cache, which specifically includes the following steps: step 1, establishing a regional user interest model; step 2, predicting the number of content requests on a base station; step 3, calculating the cache benefit of the content on the base station; step 4, establishing a cache strategy to execute the cache; step 5, updating the cache benefit of the content. The present invention calculates the cache benefit of the content based on the number of historical requests from users, the degree of matching between the content and the interests of users in the region, and the cache capacity of the current base station, and places the content in the base station with the largest cache benefit. Compared with traditional cache algorithms, this algorithm effectively reduces the latency of content transmission in the region. And the algorithm will adaptively place copies of content with higher popularity on multiple base stations in the region, effectively reducing the transmission pressure of hot content in the region and improving the user's QoE.

Description

Technical Field

[0001] The present invention belongs to the technical field of cache, and in particular relates to a method for placing content in an edge cache. Background Art

[0002] With the explosive growth of data and the deepening interconnection of various forms of things, the growth of data has far exceeded the growth of network bandwidth. Therefore, simply maintaining the original network computing method or continuously improving bandwidth cannot meet the current user requirements for network computing capabilities. In addition, technological advances in today's era have led to the emergence of many new applications, such as smart manufacturing and unmanned driving technology, all of which have put forward requirements for data transmission and processing efficiency. And the rapid development of smart terminals has brought about a sharp increase in the number of mobile devices. In order to cope with so many challenges, how to effectively reduce transmission delay and improve network transmission quality has become a highly concerned issue. The construction of content delivery network (CDN) and edge computing provide an effective solution. As a special network different from traditional networks, CDN implements content scheduling and load balancing by deploying distributed edge servers, thereby ensuring efficient content delivery. The edge caching strategy caches popular content in nearby cache base stations in advance before the network traffic reaches its peak, so that many users obtain content from edge cache base stations instead of central networks, effectively reducing the transmission consumption of duplicate content and improving the user experience.

[0003] Edge cache research is mainly divided into two parts. The first is the research on cache content, which is mainly to solve the problem of what content to cache. The second is the research on cache deployment, which is mainly to solve the problem of where to cache. According to different cache timings, cache strategies are divided into passive cache strategies and active cache strategies. The most classic passive cache strategies are the least recently used LRU algorithm and the least recently used LFU algorithm. The LRU algorithm assumes that the content requested in the recent period is more likely to be requested in the next period of time. When the cache capacity reaches the upper limit, the content that has not been requested for the longest time will be deleted; the LRU algorithm assumes that the content with a high request frequency in the recent period is more likely to be requested in the next period of time. When the cache capacity reaches the upper limit, the content with the lowest recent request frequency will be deleted. Although the passive cache strategy is relatively simple to implement, the performance of different request modes is greatly different, and it is impossible to cache popular content on the Internet to the server in advance, and it is impossible to make different optimizations for different users. The active cache strategy can use machine learning technology to predict user interest preferences, so it can cache content more targeted and improve the user experience. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a method for placing content in an edge cache that calculates the cache benefit of content based on the number of historical requests of users, the degree of matching between content and the interests of users in a region, and the cache capacity of the current base station, and places the content in the base station with the largest cache benefit, thereby effectively reducing the delay of content transmission in the region, alleviating the transmission pressure of hot content in the region, and improving the user QoE.

[0005] The objective of the present invention is achieved through the following technical solutions: a method for placing content in an edge cache, deploying multiple micro base stations SBS in a macro base station MBS to form a cache cooperation domain, wherein each SBS provides services for multiple users, and the SBSs obtain cache information of other SBSs through the MBS; dividing the user's request processing into three types: if the SBS to which the user belongs caches the content requested by the user, the SBS directly transmits the content to the user; if the SBS to which the user belongs does not cache the content requested by the user, but other SBSs in the cache cooperation domain cache the content, the SBS that caches the content and is closest to the user transmits the content to the SBS to which the user belongs, and then the SBS to which the user belongs delivers the content to the user; if the entire cache cooperation domain does not cache the content, the SBS to which the user belongs will request the content from the backbone network;

[0006] Content placement specifically includes the following steps:

[0007] Step 1: Establish regional user interest model;

[0008] Step 2: predict the number of requests for the content on the SBS;

[0009] Step 3: Calculate the caching benefit of the content on the SBS;

[0010] Step 4: Establish a cache strategy to execute cache;

[0011] Step 5: Update the cache benefits of the content.

[0012] The specific implementation method of step 1 is: process the user's historical request records, take the content requested by the user as positive samples and randomly add content that the user has not requested as negative samples, then send all samples to the deep interest network DIN model for training to obtain the user's interest model, and finally input the user's request record sequence into the model to obtain the user's interest score for each content.

[0013] The specific implementation method of step 2 is as follows: predict the request content on the SBS according to the user's historical request record: first, sort the content that each user is interested in from high to low according to the interest score obtained in step 1, then take a part of the content with higher scores to form the predicted request set of the SBS, and the number of the content taken is equal to the historical average number of requests of the user; then count the predicted request set of the SBS to obtain the predicted request number of content r placed on SBS j

[0014] The specific implementation method of step 3 is: the cache benefit refers to the average request delay reduction degree that can be brought about by placing the content on the designated server, which is determined by the number of requests and the cache capacity of the base station; by calculating the delay difference before and after caching the content, the delay benefit of placing the content r on SBS j is obtained as:

[0015]

[0016] Where N is the set of SBSs; Indicates whether SBS j has cached content r. If so, otherwise is the transmission delay between SBS k and the nearest SBS i that has cached the content;

[0017] Will Set to 1, and then recalculate the transmission delay between base stations to obtain the transmission delay t' after caching content r to SBS j ki ;

[0018] The cache capacity of the base station, that is, the remaining capacity of the base station m j Taking this into account, the final cache benefit expression is:

[0019]

[0020] The specific implementation method of step 4 is: sort the contents in the cache set according to cache benefits, cache the contents with the largest benefits in the designated SBS, and then update the cache benefits of related contents, and repeat the operation until the cache space is exhausted.

[0021] The specific implementation method of step 5 is: according to the cache revenue expression Updates the cache benefit of cached content.

[0022] The beneficial effects of the present invention are as follows: the present invention proposes a method for placing content in an edge cache that can achieve lower transmission delays. The method calculates the cache revenue of the content based on the number of historical requests from users, the degree of matching between the content and the interests of users in the region, and the cache capacity of the current base station, and places the content in the base station with the largest cache revenue. Compared with traditional cache algorithms, this algorithm effectively reduces the latency of content transmission in the region. In addition, the algorithm will adaptively place copies of content with higher popularity in multiple base stations in the region, effectively reducing the transmission pressure of hot content in the region and improving the user's QoE. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A scene architecture diagram of a specific embodiment of the present invention;

[0024] Figure 2 The figure is a flowchart of an algorithm according to a specific embodiment of the present invention. DETAILED DESCRIPTION

[0025] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0026] The content placement method in the edge cache of the present invention calculates the cache benefit of the content based on the number of historical requests of users, the degree of matching between the content and the interests of users in the area, and the cache capacity of the current base station, and places the content in the base station with the largest cache benefit. Applicable scenarios include Figure 1 As shown: multiple micro base stations SBS are deployed in the macro base station MBS to form a cache cooperation domain, where each SBS provides services for multiple users, and SBSs obtain cache information of other SBSs through MBS; user request processing is divided into three types: if the SBS to which the user belongs caches the content requested by the user, the SBS directly transmits the content to the user; if the SBS to which the user belongs does not cache the content requested by the user, but other SBSs in the cache cooperation domain cache the content, the SBS that caches the content and is closest to the user transmits the content to the SBS to which the user belongs, and then the SBS to which the user belongs delivers the content to the user; if the entire cache cooperation domain does not cache the content, the SBS to which the user belongs will request the content from the backbone network;

[0027] Compared with the traditional caching algorithm, the invention effectively reduces the delay of content transmission in the region. And the algorithm will adaptively place copies of content with higher popularity in multiple base stations in the region, effectively reducing the transmission pressure of hot content in the region and improving the user's QoE (Quality of Experience). Its working principle is: the base station is responsible for collecting the user's request records, and using machine learning methods to analyze and train the user's requests to obtain the user's interest model. The number of content requests reflects the popularity of the content in the current base station. Content with a large number of requests will show a greater delay benefit. The number of content requests can be predicted by the similarity between the content and the user's interest and the user's historical average number of requests. From each user's sorted set of interested content, the content with the same number as the user's historical average number of requests is taken out and spliced ​​into the request set of the base station, and the predicted number of requests for each content is obtained by statistics. Delay benefit refers to the difference between the total request delay before and after the content is cached to the base station. The base station with a large number of requests and a physical location close to the center of the region will obtain a larger delay benefit. However, if only the delay benefit is used as the cache benefit, it will lead to a center effect, that is, the base station geographically close to the center of the region will cache more content than other base stations. To this end, the cache capacity of the base station, that is, the remaining capacity of the base station m i Taking this into account, we get the final cache benefit expression. Finally, we place the content based on the cache benefit. Each time, we take out the content with the largest cache benefit and place it in the corresponding base station. We update the cache benefit of the relevant content and repeat this step until the cache capacity of all base stations reaches the upper limit.

[0028] like Figure 2 As shown, content placement specifically includes the following steps:

[0029] Step 1, establish a regional user interest model; the specific implementation method is: process the user's historical request records, take the content requested by the user as positive samples and randomly add the content that the user has not requested as negative samples, and then send all samples to the deep interest network DIN model for training to obtain the user's interest model, and finally input the user request record sequence into the model to obtain the user's interest score for each content.

[0030] Step 2: Predict the number of requests for content on the SBS; considering that different users have different request capabilities, the present invention predicts the request content of the base station based on the user's historical request number. The specific implementation method is: predict the request content on the SBS based on the user's historical request record: first, sort the content that each user is interested in from high to low according to the interest score obtained in step 1, then take a portion of the content with higher scores to form the predicted request set of the SBS, and the number of content taken is equal to the user's historical average request number; then count the predicted request set of the SBS to obtain the predicted request number of content r placed on SBS j

[0031] Step 3: Calculate the caching benefit of the content on the SBS. The caching benefit refers to the average request delay reduction that can be achieved by placing the content on a specified server, which is determined by the number of requests and the caching capacity of the base station. By calculating the delay difference before and after caching the content, the delay benefit of placing content r on SBS j is:

[0032]

[0033] Where N is the set of SBSs; Indicates whether SBS j has cached content r. If so, otherwise is the transmission delay between SBS k and the nearest SBS i that has cached the content;

[0034] Will Set to 1, and then recalculate the transmission delay between base stations to obtain the transmission delay t' after caching content r to SBS j ki ;

[0035] However, if only latency benefit is used as cache benefit, it will lead to the center effect, that is, the base station located near the center of the region will cache more content than other base stations. Therefore, the cache capacity of the base station, that is, the remaining capacity of the base station m j Taking this into account, the final cache benefit expression is:

[0036]

[0037] Step 4, establish a cache strategy to execute cache; the specific implementation method is: take the cache cooperation domain composed of multiple SBSs deployed in the MBS as the object, minimize the average request delay of the content as the goal, and consider the interest differences between base stations and the popularity of the content, formulate a joint cache strategy in the region, and cache the content according to the size of the cache benefit; sort the content in the cache collection according to the cache benefit, cache the content with the largest benefit in the designated SBS, and then update the cache benefit of the related content, and repeat the operation until the cache space is exhausted.

[0038] Step 5: Update the cache revenue of the content; the specific implementation method is: according to the cache revenue expression When the cache situation in the region changes, the cache benefit will also change. Therefore, after caching the content with the largest benefit, the cache benefit of the cached content needs to be updated. The cache benefit of the cached content in other cache base stations will decrease, but for the content with a large number of requests, its cache benefit is likely to still be higher than that of unpopular content, so it will be cached repeatedly in the region to cope with the high number of requests.

[0039] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A method for placing content in an edge cache, characterized in that: Multiple micro base stations (SBSs) are deployed in a macro base station (MBS) to form a cache cooperation domain, where each SBS provides services to multiple users, and SBSs obtain cache information of other SBSs through MBSs. User request processing is divided into three types: if the SBS to which the user belongs has cached the content requested by the user, the SBS directly transmits the content to the user; if the SBS to which the user belongs does not cache the content requested by the user, but other SBSs in the cache cooperation domain have cached the content, the SBS that has cached the content and is closest to the user transmits the content to the SBS to which the user belongs, and the SBS to which the user belongs then delivers the content to the user; If the entire cache cooperation domain does not cache the content, the SBS to which the user belongs will request the content from the backbone network; Content placement specifically includes the following steps: Step 1: Establish a regional user interest model; the specific implementation method is: process the user's historical request records, take the content requested by the user as positive samples and randomly add the content that the user has not requested as negative samples, then send all samples to the deep interest network DIN model for training to obtain the user's interest model, and finally input the user request record sequence into the model to obtain the user's interest score for each content; Step 2: Predict the number of requests for content on the SBS. The specific implementation method is: predict the requested content on the SBS based on the user's historical request records: first, sort the content that each user is interested in from high to low according to the interest score obtained in step 1, then take a portion of the content with higher scores to form the predicted request set of the SBS, and the number of the content taken is equal to the user's historical average request number; then count the predicted request set of the SBS to obtain the predicted request number of content r placed on SBS j Step 3: Calculate the caching benefit of the content on the SBS. The specific implementation method is: the caching benefit refers to the average request delay reduction degree that can be brought about by placing the content on a designated server, which is determined by the number of requests and the caching capacity of the base station. By calculating the delay difference before and after caching the content, the delay benefit of placing content r on SBS j is obtained as: Where N is the set of SBSs; Indicates whether SBS j has cached content r. If so, otherwise is the transmission delay between SBS k and the nearest SBSi that has cached the content; Will Set to 1, and then recalculate the transmission delay between base stations to obtain the transmission delay t' after caching content r to SBS j ki ; The cache capacity of the base station, that is, the remaining capacity of the base station m j Taking this into account, the final cache benefit expression is: Step 4: Establish a cache strategy to execute cache. The specific implementation method is: sort the contents in the cache set according to cache benefits, cache the contents with the largest benefits in the specified SBS, and then update the cache benefits of the relevant contents, and repeat the operation until the cache space is exhausted; Step 5: Update the cache revenue of the content; the specific implementation method is: according to the cache revenue expression Updates the cache benefit of cached content.

Images