CN107911711B - Edge cache replacement improvement method considering partitions - Google Patents

Edge cache replacement improvement method considering partitions Download PDF

Info

Publication number
CN107911711B
CN107911711B CN201710996529.7A CN201710996529A CN107911711B CN 107911711 B CN107911711 B CN 107911711B CN 201710996529 A CN201710996529 A CN 201710996529A CN 107911711 B CN107911711 B CN 107911711B
Authority
CN
China
Prior art keywords
video
popularity
cache
file
time
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201710996529.7A
Other languages
Chinese (zh)
Other versions
CN107911711A (en
Inventor
张鹤立
刘洪燕
李曦
纪红
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing University of Posts and Telecommunications
Original Assignee
Beijing University of Posts and Telecommunications
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing University of Posts and Telecommunications filed Critical Beijing University of Posts and Telecommunications
Priority to CN201710996529.7A priority Critical patent/CN107911711B/en
Publication of CN107911711A publication Critical patent/CN107911711A/en
Application granted granted Critical
Publication of CN107911711B publication Critical patent/CN107911711B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
    • H04N21/23—Processing of content or additional data; Elementary server operations; Server middleware
    • H04N21/231—Content storage operation, e.g. caching movies for short term storage, replicating data over plural servers, prioritizing data for deletion
    • H04N21/23106—Content storage operation, e.g. caching movies for short term storage, replicating data over plural servers, prioritizing data for deletion involving caching operations
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
    • H04N21/23—Processing of content or additional data; Elementary server operations; Server middleware
    • H04N21/231—Content storage operation, e.g. caching movies for short term storage, replicating data over plural servers, prioritizing data for deletion
    • H04N21/23113—Content storage operation, e.g. caching movies for short term storage, replicating data over plural servers, prioritizing data for deletion involving housekeeping operations for stored content, e.g. prioritizing content for deletion because of storage space restrictions
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
    • H04N21/23—Processing of content or additional data; Elementary server operations; Server middleware
    • H04N21/24—Monitoring of processes or resources, e.g. monitoring of server load, available bandwidth, upstream requests
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
    • H04N21/23—Processing of content or additional data; Elementary server operations; Server middleware
    • H04N21/24—Monitoring of processes or resources, e.g. monitoring of server load, available bandwidth, upstream requests
    • H04N21/2408—Monitoring of the upstream path of the transmission network, e.g. client requests
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
    • H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
    • H04N21/251—Learning process for intelligent management, e.g. learning user preferences for recommending movies
    • H04N21/252—Processing of multiple end-users' preferences to derive collaborative data
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
    • H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
    • H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
    • H04N21/258—Client or end-user data management, e.g. managing client capabilities, user preferences or demographics, processing of multiple end-users preferences to derive collaborative data
    • H04N21/25866—Management of end-user data
    • H04N21/25891—Management of end-user data being end-user preferences
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W28/00—Network traffic management; Network resource management
    • H04W28/02—Traffic management, e.g. flow control or congestion control
    • H04W28/10—Flow control between communication endpoints
    • H04W28/14—Flow control between communication endpoints using intermediate storage

Abstract

The invention discloses an edge cache replacement improvement method considering partitions, and belongs to the technical field of wireless communication. The method comprises the steps that firstly, a cache region C is divided into two regions, namely a new resource cache region C1 and an old resource cache region C2; then, the popularity of the new video is estimated before the user accesses, and the actual popularity of the video is updated when the videos in C1 and C2 are accessed once; finally, replacing the video in the buffer area, and enabling the buffering time in the C1 to exceed TsMoves out to C2, when the actual popularity of a certain video in C1 is the lowest in the buffer C, the video is preferentially replaced; if the video content is not in the cache C and is higher than the lowest popularity of the cache C, if no space is left for storage, the files with the lowest actual popularity in the cache area are sequentially replaced until enough storage space is available. The invention makes the cell base station obtain the optimal caching benefit, adapts to the rapid change of the popularity of the network access resource and makes the network performance more stable.

Description

Edge cache replacement improvement method considering partitions
Technical Field
The invention belongs to the technical field of wireless communication, and particularly relates to an edge cache replacement improvement method considering partitioning.
Background
According to the prediction of the Cisco virtual network index, in 2018, various forms of multimedia traffic can reach 90% of network traffic (reference 1: Cisco, "Cisco visual network index: Global mobile data for implementation update, 2012-2017," White Paper, [ Online ] http:// go. gl/uQ0DJQ,2013.), along with the rapid increase of uplink and downlink rates of wireless access, the globalization popularization of intelligent terminals and the trend of network video traffic showing exponential increase under the Global environment of continuous development of media. Although the development of wireless communication technology has been very rapid in recent years, today's 4G access networks can provide bandwidths up to 100Mbps and 1Gbps for mobile users and stationary users, respectively. However, the development of the mobile core network seems to encounter a bottleneck, and the link bandwidth from the base station to the Internet via the mobile center network is far from meeting the requirement of the user. In order to reduce the risk probability that the network is exposed to link congestion, the operator may use caching technology to shorten the distance between the user and the requested content. Therefore, in radio access networks, applying caching is considered to be the fastest and most efficient solution today. However, the cache space is limited, the popularity of the content requested by the user is also changing continuously, and after the cache space is full, the data which no longer has the cache value in the cache needs to be released according to a certain replacement rule, so that the space is made free to cache the new data which has the cache value, and the request hit rate and the byte hit rate of the cache space are improved.
Existing cache replacement algorithms are mainly classified into four types: based on access object size, based on access content time interval, based on access content access times, based on access content value, respectively. The replacement algorithms only consider the access characteristics of the accessed content and ignore the most important factor, namely a network terminal user, so that the cache hit rate is not ideal. The access behavior of the network terminal user determines various access characteristics of the accessed content, the access behavior of the user has uncertainty, the uncertainty causes the popularity of the accessed content to be in continuous change, and how to more accurately grasp the popularity of the network content is a key factor for improving the cache hit rate.
According to 2016 Chinese network audio-visual development research reports, the reported investigation results show that by 2016, 6 months, the scale of Chinese network video users reaches 5.14 hundred million, and the occupation ratio of mobile users watching Internet videos for more than 5 days per week is as high as 69.9%. The research report from the CNNIC network video user shows that the types of the most popular programs among the long videos are movies, TV shows and comprehensive art, which account for 50.9%, 48.8% and 33.7% respectively, and the types of the most popular programs among the short videos are news information, funny videos and entertainment videos, which account for 30.0%, 25.4% and 18.2% respectively. According to big data statistics, only a small part of videos are frequently accessed by the user within a period of time, the two-eight law is highlighted, and the user has more preference for new network contents. Therefore, the video with the higher updating speed has the higher popularity change speed, and if the updated video has the relevance, the popularity also has the relevance. By taking these factors into account, the present invention will propose an improved cache replacement method in a mobile network.
The most important factor in the cache replacement algorithm is the popularity index, and if the popularity index is determined, the cache replacement algorithm is determined, and the least popular cache in the cache area is replaced. The classical cache replacement algorithm evaluates the popularity of a network resource, and usually adopts the indexes of time and quantity, for example, the passive cache replacement algorithms adopted in the current network are mainly lru (least recent used) and lfu (least frequent used) algorithms. The LRU algorithm defines popularity by using the interval between the current access time and the last access as an index, i.e., by using a rule that the shorter the interval is, the higher the popularity is. The LFU algorithm defines the popularity of the current content based on recording the total number of times the network content is accessed by the user from the beginning of production to the current time, i.e., with rules that increase popularity with higher number of accesses. Researchers have studied a Content popularity-based mpv (most temporal video) replacement algorithm, which requires a very large cache space to obtain a high cache benefit and is mostly applied to a CDN (Content Delivery Network).
The existing replacement algorithm has limitations on the algorithm itself, for example, the LRU algorithm considers that the least recently used content is replaced, and is very easily interfered by access noise, which is quite serious, and the performance of the algorithm is not ideal. The LFU algorithm, which considers the replacement of the content used least frequently for a long time, has a drawback in that it cannot timely eliminate the content whose access volume has been increased rapidly in the past short time and which has no cache value thereafter. The FIFO replacement algorithm not only causes resource waste due to frequent replacement, but also cannot prolong the retention time of the resource with high popularity in the cache region, resulting in low cache hit rate. In addition, some cache replacement policies increase the number of requests served by cached content by prolonging the time that some content is in the cache space, and this does not stably guarantee that the cache value of those content resources whose cache time is prolonged is higher than that of the content resources that are not cached. In summary, except for the defects of the self algorithm, most of the existing cache replacement algorithms do not consider the behavior characteristics of the users from the perspective of the users, and the influence of the difference of popularity trends of different kinds of network contents on the cache is ignored.
Disclosure of Invention
In order to improve the hit rate of the cache in the mobile network to improve the network performance, the invention analyzes the factor of the unsatisfactory hit rate caused by the existing cache replacement algorithm, and provides an edge cache replacement improvement method considering the partition.
The invention provides an edge cache replacement improvement method considering partitioning, which comprises the following steps:
step 1, dividing a cache region C into two regions, namely a new resource cache region C1 and an old resource cache region C2;
the video file stored at C1 needs to satisfy three conditions: firstly, the new video content generated in the network, secondly, the estimated popularity of the new video is higher than the lowest popularity of the video in the C, and thirdly, the residence time in the C1 does not exceed the preset time Ts(ii) a When the video buffering time exceeds TsThen the video will be considered old;
c2 storing video according to its actual popularity and the video files have been in the network for more than Ts；
Step 2, estimating the popularity of the new video before the user accesses, and updating the actual popularity of the video when the videos in C1 and C2 are accessed once;
only when the new video has the relevant continuous set in the cache region C, setting the estimated popularity of the new video to be equal to the actual popularity of the video file of the relevant continuous set in the cache region C;
and 3, replacing the video in the cache region according to the following strategy:
(1) the buffering time in C1 exceeds TsMove out to C2;
(2) when the actual popularity of a certain video in the C1 is the lowest in the cache C, the video is preferentially replaced;
(3) when the video content accessed by the user is not in the cache C, calculating the actual popularity of the video, if the actual popularity is higher than the lowest popularity of the cache C, further judging whether the cache area has residual space for storage, if so, storing the video, and if not, sequentially replacing the files with the lowest actual popularity in the cache area until enough storage space exists; if the video is a newly generated video file in the network, it is cached in C1, otherwise it is cached in C2.
The invention has the advantages and positive effects that: the invention provides an improved edge cache replacement method under the structure of a micro-cellular mobile network, which combines the current network development trend and the user behavior characteristics of mobile network users, ensures that a cell base station utilizes limited cache space to exchange optimal cache benefit by improving the hit rate of the edge cache, can adapt to the rapid change of the popularity of network access resources in real time, and ensures that the network performance becomes more stable.
Drawings
FIG. 1 is a schematic diagram of a mobile network edge caching model;
FIG. 2 is a schematic overall flow chart of an improved method for edge cache replacement according to the present invention;
FIG. 3 is a comparison graph of the caching performance of the method of the present invention and two existing caching algorithms in different caching spaces;
FIG. 4 is a comparison graph of cache replacement times of the present invention method and two existing cache algorithms in different cache spaces;
FIG. 5 is a diagram showing the effect of the caching performance of the method of the present invention and two existing caching algorithms under different popularity variation spans.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples.
According to the method, firstly, according to the behavior characteristics of a user, the preference of the user to new network video resources is higher than that of old network video resources, the popularity of the new network video resources is higher in speed, the popularity of the old network video resources is lower in speed, and the popularity of the old network video resources is lower in speed, so that different popularity indexes need to be established for the new and old network video resources to reduce errors between the calculated popularity and the actual popularity, an active cache replacement strategy is adopted for the new network video resources, a passive cache replacement strategy is adopted for the old network video resources, and the situations that access noise interference and transient high-access videos occupy cache space for a long time can be avoided. And secondly, the popularity of the new network video resource is estimated by introducing the content relevancy. The following detailed description is made with reference to the accompanying drawings and examples.
As shown in fig. 1, the network scenario is that Small cell base stations (Small cells) all have the capability of edge caching, and store local hotspot videos into a Small base station Cache in a local area, when a Small cell user requests content, if the Small base station has cached corresponding content, the Small base station directly provides service for the user, and if the Small base station does not have the content, the Small base station obtains the content from a content server in the Internet through a core network linked by a backhaul link from a macro base station. In the figure, the sgw (serving gateway) is a serving gateway, and the pgw (packet Data Network gateway) is a packet Data gateway. Because the cache capacity is limited, the hot content in the network changes at any time, so the content cached in the small base station needs to be replaced in time to meet the requirement of serving the user to the maximum extent. Based on the research, the invention provides an improved edge cache replacement method based on video, because video traffic is dominant in a mobile network.
Fig. 2 is a flow chart of the improved method for edge cache replacement according to the present invention, and the following steps are described.
Step 1) partitioning the cache space.
The invention divides the buffer space into a new resource buffer C1 and an old resource buffer C2, and the sum of the two buffer spaces is the total buffer capacity C. The sizes of the two buffers can be adjusted in a self-adaptive mode.
The contents stored in the C1 cache region satisfy three conditions: firstly, the new video content generated in the network, and secondly, the estimated popularity of the new video is higher than the minimum popularity p of the video file in the cache region CminThirdly, the time for buffering in the C1 area cannot exceed the time TsWhen the buffering time of the content of the C1 buffer exceeds TsIt is transferred to the C2 cache. T issThe time length is preset, and after the time length is exceeded, the new video is marked as the old video.
The condition that the video files in the C2 cache area are all satisfied is that the video files exist in the network for more than T times. C2 caches the video files according to their actual popularity, the higher the actual popularity of the video is, the better the video is stored in the buffer C2.
And 2) when a new video content is generated, estimating the popularity of the new video before the user accesses the new video content. The popularity of the video is updated each time the video in C1 and C2 is accessed.
The popularity of a new movie or television show or art will often have a strong relationship with previous promotions, the enthusiasm of the actors, the genre of the show, the production team, etc. These three types of videos are also the most preferred video file types for video users. And rough popularity prediction can be made according to the preference of user groups to different types of video files, and as big data is continuously developed, popularity prediction errors are continuously reduced. In fact, for the video of each episode, the popularity can often be estimated by the actual popularity of the last episode associated therewith, i.e. the popularity of the associated video episode has a continuation, such as a tv series, a full art week show, a movie series, etc. Therefore, the invention introduces the video content relevance r, and in order to simplify the algorithm, when a new video file is a relevant continuous set of video files in the cache region C, the estimated popularity value of the new video file is equal to the actual popularity value of the relevant video file in the cache region C by setting r to be 1. If no video file is related to the new video file in the buffer C, r may be set to 0, the default setting may estimate the popularity to be low, for example, 0, and at this time, the video may not be directly stored in the buffer C1, so that it may be determined whether to store the video in the buffer in time according to the popularity calculation value after the actual user accesses the video.
Before a user accesses, popularity estimation is carried out on new video content, and two situations exist: firstly, when the popularity pre-estimated value is higher than the minimum popularity value of the cache region, the popularity pre-estimated value is pre-stored in the C1 cache region and is stored in the C1 cache region, then the actual popularity is calculated according to the actual access times of the user, and if the actual popularity is lower than the minimum popularity p of the cache regionminIf so, it is deleted, otherwise, it is cached in the C1 cache until the residence time T is exceededsThen moves into the C2 buffer area; and secondly, when the estimated value is lower than the minimum popularity value of the cache region, the caching operation is not carried out, but when the actual access frequency of the user to the new video content is higher than the minimum popularity value of the cache region, the new video content needs to be cached in the C1 cache region in time.
The actual popularity of the video is updated each time the video in C1 and C2 is accessed.
The actual popularity p of the video file in the cache area of C1 is calculated by the following formula:
Figure BDA0001442614410000051
wherein R istIndicating the number of times the video file was accessed in each statistical window. s denotes the size of the video file. beg denotes the start of the statistical period and cur denotes the expiration of the statistical period. The statistical window refers to the popularity statistics over the same time interval.
The calculation of the actual popularity of the video file in the C2 cache area is calculated by the EWMA (organization weight moving-Average, control chart) algorithm as follows:
Tn+1＝(1-α)*Tn+α*(tcur-tlast) (2)
p＝1/Tn+1/s (3)
wherein, Tn+1Indicating the average access time interval, t, of the currently accessed video filecurIndicating the current access time, t, of the video filelastThe time indicating the last access of the file is α, which is a weighting factor, and the value is decimal between 0 and 1, and it indicates the influence of the current access on the current popularity and also indicates the forgetting degree on the historical popularity.
And 3) replacing the video in the cache region.
First, the size of the cache space of the cache region C1 and the cache region C2 is not fixed, and can be adaptively adjusted to maximize the cache hit rate. Secondly, each time a user accesses a video file, the actual popularity of the video file in the cache area changes.
The following conditions can cause the content state of the file in the cache area to change:
once a video in the buffer C1 is the lowest in the buffer C when it is actually popular, the video content will be replaced preferentially.
Secondly, when the buffering time of the video in the buffer C1 exceeds the residence time TsAnd then move to buffer C2.
And thirdly, when the video content accessed by the user is originally not in the cache regions C1 and C2, but the access popularity of the video content is higher than the actual popularity of part of the video files in the cache regions, deleting part of the files with the lowest cache popularity in the cache regions, and making room for caching the video files. If the file is a newly generated video file in the network, the file is cached in a C1 cache area in time; if the file is not a new video file, the file is directly cached in the cache area C2.
As shown in FIG. 3, the buffer C1 will buffer for a period of time exceeding T at any timesThe video is shifted out. When receiving user request, searching video to be requestedIf u is in the cache region C, updating the record item information of the video u, updating and calculating the actual popularity of the video u, and transmitting the cached video data to the user. And if the video u requested by the user is not in the buffer C, inquiring whether the buffer C has the residual space, if so, creating a record item for the video u in the buffer, and caching the video u. If the buffer C has no residual space, calculating the popularity of the video u, and when the popularity of the video u is higher than the lowest popularity p of the videos in the buffer CminAnd deleting the video record item with the lowest popularity in the cache region C, creating a record item for the video u in the cache region, and caching the video u. When the video file u is cached, the video file data is obtained from the upper network node, if the video is a new video generated in the network, the video is cached in the C1, and if not, the video is cached in the C2.
In order to cache more efficiently, the access information and the attribute of the relevant file need to be recorded, so that the popularity calculation and the cache area content management are facilitated. It needs to occupy 34 bytes of storage space, and for a network file, the occupied space is not very large, especially for a large video file, it is worth sacrificing the buffer space.
The contents of the fields contained in the cache entry are listed below
Table 1 entry information of edge cache file
Field symbol Data type Occupied byte number Physical meanings
tlast Shaping machine 4 Last access time of file
tcur Shaping machine 4 Current access time of file
tbeg Shaping machine 4 File recording start time
r Shaping machine 1 Document relevancy
p Long floating point type 8 Current actual popularity
location Shaping machine 4 Location of file cache
class Shaping machine 1 Buffer area number
count Shaping machine 4 Number of accesses
size Shaping machine 2 File size
α Floating point 2 EWMA algorithm popularity weighting factor
The entry is updated each time the video is accessed and the popularity of the video file is recalculated. The minimum popularity of the video in the buffer C needs to be updated in time according to the actual popularity of the video.
Example (b):
the simulation assumes that N network video files can be accessed by a user currently, and X new video files are generated every day, so that in order to simplify the cache replacement process, according to the favorite video file types of the user, namely movies, television shows, news information, and funny videos, the size of each video file in the simulation is one of four values of 400M, 200M, 100M and 30M, and the value of N is set to 10000 in the simulation. The value of X is set to 50 and the buffer space size is C.
The probability that each file in the network is accessed by the user is different, the access probability of the file approximately has the rule of Zipf distribution, namely the video file with the first popularity is accessed with the highest probability, the lower the popularity is, the lower the access probability is, the trend of exponential decay is presented, and the access probability p of the ith file is distributed according to the ZipfiComprises the following steps:
Figure BDA0001442614410000061
where N is the aforementioned number of all video files currently accessible to the user, αzThe value of the parameter of Zipf distribution is between 0 and 1, the size of the parameter determines the speed of the access probability decreasing along with the ranking, the access is more concentrated if the value is larger, α is obtained in the simulation according to the twenty-eight lawzThe value of (d) is set to 0.8.
However, since there is random variability in the access of the user to the network video and there is a certain error in the popularity calculation, the embodiment assumes that there are e error bits in the popularity ranking, so the file with the ith popularity ranking in the simulation will adopt the following access probability:
Figure BDA0001442614410000071
wherein random (a, b) represents a random integer between a and b (excluding a, b).
In the simulation, the time interval of the user access request obeys the poisson distribution, lambda is used for representing the mean value of the poisson distribution, namely the average access request interval, the lambda value is set to be 60 seconds, the random number of the user access time interval can be generated through the poisson distribution, the file accessed by the request is determined through the approximate Zipf distribution of the formula (5), and then one access event is generated. The simulation experiment generates 100 ten thousand access events, and the following table 2 lists parameters used in the simulation, and parameter default values and parameter meanings.
The detailed simulation parameters are shown in table 2:
table 2 simulation parameter settings
Parameter field Meaning and default value of parameter
N Number of network video files (10000) that the current user can access
X Number of network video files newly added every day (50)
λ User visit average time interval (60 seconds)
αz Popularity Zipf distribution parameter (0.8)
C Buffer memory space size (40000M)
C1 Space size of C1 buffer (dynamic size, C1+ C2 ═ C)
C2 Space size of C2 buffer (dynamic size, C1+ C2 ═ C)
e Popularity error digit (5)
α EWMA Algorithm weighting factor (0.6)
s Video file size (400M, 200M, 100M, 30M)
Ts C1 buffer term (48h)
Hit Ratio Cache hit rate
The embodiment performs simulation experiments on the improved caching method provided by the invention from different angles, and compares the improved caching method with the traditional LRU and LFU algorithms. On the one hand, the simulation comparison of the cache hit rate and the cache replacement times under different cache space sizes, and on the other hand, the representation of the cache hit rate under different popularity change speeds (represented by the number of files of the popularity change span).
Firstly, the method of the present invention is marked as an improved replacement algorithm in fig. 3 to 5, compared with the conventional two cache replacement algorithms of LRU and LFU, performance comparison simulation is performed under different cache space sizes, and the simulation result is shown in fig. 3. On the other hand, the instability of the cache performance caused by the time limitation of the conventional replacement algorithm is seriously influenced by the access noise, so that the cache efficiency of the conventional replacement algorithm is low when the cache space is small.
Secondly, as another factor influencing the edge cache performance is the replacement frequency of the cache, the less the replacement frequency of the cache is, the less the occupied network resources are, and the utilization rate of the network resources can be effectively improved. The simulation comparison graph is shown in fig. 4, the number of times of user access in simulation is 1 ten thousand, and the popularity still obeys the Zipf distribution with the parameter of 0.8, and it can be seen from the graph that the number of times of cache replacement is obviously much less than that of the traditional LRU replacement algorithm in any proportion of cache space, and is also obviously superior to the traditional LFU algorithm after the cache space is more than ten percent.
Finally, starting from the change speed of the popularity of the network video resources, the span of the popularity ranking change is a variable, and the result of comparing the improved replacement algorithm scheme with the traditional replacement algorithm scheme through simulation is shown in fig. 5.
The invention combines the larger difference of the new and old network resources on the popularity trend to formulate different popularity indexes, avoids the situation that the transient high-access new video still occupies the cache space for a long time after a period of time, and simultaneously, also combines the content continuity of the network video resources and introduces the content relevance to estimate the popularity of the continuous content. By classifying and caching new and old network resources, the new network resources calculate popularity according to the content relevancy to perform active cache replacement, and the old network resources adopt passive cache replacement according to the actual popularity. Through verification, the invention can improve the hit rate of the edge cache, so that the cell base station utilizes the limited cache space to obtain the optimal cache benefit and adapts to the rapid change of the popularity of the network access resource in real time.

Claims (6)

1. An improved method for partition-based edge cache replacement, comprising the steps of:
step 1, dividing a cache region C into two regions, namely a new resource cache region C1 and an old resource cache region C2;
the video file stored at C1 needs to satisfy three conditions: firstly, the new video content generated in the network, and secondly, the estimated popularity of the new video is higher than that of the C middle viewLowest prevalence of frequency, three is that the residence time in C1 does not exceed the preset time Ts(ii) a When the video buffering time exceeds TsThen the video will be considered old;
c2 storing video according to its actual popularity and the video files have been in the network for more than Ts；
Step 2, estimating the popularity of the new video before the user accesses, and updating the actual popularity of the video when the videos in C1 and C2 are accessed once;
when the new video file is the relevant continued set of the video file in the cache region C, the estimated popularity value of the new video file is equal to the actual popularity value of the relevant video file in the cache region C;
and 3, replacing the video in the cache region according to the following strategy:
(1) the buffering time in C1 exceeds TsMove out to C2;
(2) when the actual popularity of a certain video in the C1 is the lowest in the cache C, the video is preferentially replaced;
(3) when the video content accessed by the user is not in the cache C, inquiring whether the buffer C has a residual space or not, if so, caching the video, if the buffer C has no residual space, calculating the popularity of the video, and when the popularity of the video is higher than the lowest popularity of the video in the buffer C, deleting the video with the lowest popularity in the buffer C and caching the video; if the video is a newly generated video file in the network, it is cached in C1, otherwise it is cached in C2.
2. The improvement method of edge cache replacement considering partitions as claimed in claim 1, wherein the size of the cache space of C1 and C2 can be adaptively adjusted, and the sum of the space of C1 and C2 is the total capacity of the cache region C.
3. The partition considered edge cache replacement improvement method according to claim 1, wherein in said step 2, every time the video in C1 is accessed,updating actual popularity of video
Figure FDA0002437986300000011
Where s denotes the size of the video file, beg denotes the start of the statistical period, cur denotes the end of the statistical period, RtRepresenting the number of times the video file in each statistical window is accessed;
every time the video in C2 is accessed, the actual popularity p of the video is updated according to:
Tn+1＝(1-α)*Tn+α*(tcur-tlast)；p＝1/Tn+1/s；
wherein, Tn+1Representing an average access time interval of the currently accessed video file; t is tcurRepresenting the current access time of the video file; t is tlastα is a weighting factor, the value is between 0 and 1, α represents the influence of the current popularity by the current access and also represents the forgetting degree of the historical popularity.
4. The partition-considered edge cache replacement improvement method according to claim 1, wherein the method establishes a record item for each video in the cache area, the record item comprises the actual popularity and the storage location of the video, and the minimum popularity of the video in the cache area C is updated in time according to the actual popularity of the video.
5. The partition aware edge cache replacement improvement method of claim 4, wherein the video entry comprises: last access time t of filelastCurrent access time t of filecurFile recording start time tbegFile relevance r, current actual popularity p, location of file cache, cache number class, number of accesses count, file size, EWMA algorithm popularity weighting factor α.
6. The partition considered edge cache replacement improvement method of claim 5, wherein said file relevance r is set to 1 when the new video file is a relevant continuation of the video file in the cache C.
CN201710996529.7A 2017-10-24 2017-10-24 Edge cache replacement improvement method considering partitions Active CN107911711B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710996529.7A CN107911711B (en) 2017-10-24 2017-10-24 Edge cache replacement improvement method considering partitions

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710996529.7A CN107911711B (en) 2017-10-24 2017-10-24 Edge cache replacement improvement method considering partitions

Publications (2)

Publication Number Publication Date
CN107911711A CN107911711A (en) 2018-04-13
CN107911711B true CN107911711B (en) 2020-09-01

Family

ID=61841570

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710996529.7A Active CN107911711B (en) 2017-10-24 2017-10-24 Edge cache replacement improvement method considering partitions

Country Status (1)

Country Link
CN (1) CN107911711B (en)

Families Citing this family (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109344092B (en) * 2018-09-11 2023-06-23 天津易华录信息技术有限公司 Method and system for improving cold storage data reading speed
CN109271285B (en) * 2018-11-12 2022-05-27 上海碳蓝网络科技有限公司 File backup method, equipment and computer storage medium
CN111225267B (en) * 2018-11-26 2022-05-06 中国电信股份有限公司 Content cache scheduling method, device and system and content distribution network node
CN109753505B (en) * 2018-12-26 2022-06-24 济南银华信息技术有限公司 Method and system for creating temporary storage unit in big data storage system
CN109451344A (en) * 2018-12-28 2019-03-08 扬州大学 Based on the buffer replacing method of correlation rule in P2P stream media system
CN110213627B (en) * 2019-04-23 2020-09-01 武汉理工大学 Streaming media cache allocation method based on multi-cell user mobility
CN110519394B (en) * 2019-09-12 2022-02-15 广东工业大学 Information cache replacement method, device, terminal and storage medium
CN110933140B (en) * 2019-11-05 2021-12-24 北京字节跳动网络技术有限公司 CDN storage allocation method, system and electronic equipment
CN110913239B (en) * 2019-11-12 2021-03-02 西安交通大学 Video cache updating method for refined mobile edge calculation
CN111225023B (en) * 2019-11-19 2022-02-25 中国联合网络通信集团有限公司 Caching method and device
CN113225672B (en) * 2021-04-22 2022-01-28 湖南师范大学 Base station selection method supporting mobile user
CN113177069A (en) * 2021-05-08 2021-07-27 中国科学院声学研究所 Cache and query system and query method
CN113329065B (en) * 2021-05-18 2022-08-23 武汉联影医疗科技有限公司 Resource preheating method and device, computer equipment and storage medium
CN115002138B (en) * 2022-03-04 2023-07-04 昆明理工大学 Edge caching method based on cruise video data of terminal unmanned aerial vehicle
CN114640861B (en) * 2022-03-14 2023-12-12 上海小潮缇姆信息科技有限公司 Edge cache optimization method based on video popularity priori prediction

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102511043A (en) * 2011-11-26 2012-06-20 华为技术有限公司 Method for replacing cache files, device and system thereof
CN103237068A (en) * 2013-04-17 2013-08-07 北京科技大学 Content-attribute-distinguishable streaming media cache replacement method for CDN-P2P (content delivery network and peer-to-peer)
CN104469502A (en) * 2014-11-11 2015-03-25 中原工学院 Cache replacement policy based on video similarity in relevance video-on-demand system
CN104822068A (en) * 2015-04-29 2015-08-05 四达时代通讯网络技术有限公司 Streaming media proxy cache replacing method and device
CN104967861A (en) * 2015-05-27 2015-10-07 上海美琦浦悦通讯科技有限公司 CDN video buffer system and method

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8665281B2 (en) * 2008-02-07 2014-03-04 Microsoft Corporation Buffer management for real-time streaming
CN101778273A (en) * 2010-01-20 2010-07-14 中兴通讯股份有限公司 Method and device for transcribing multimedia radio programs
CN105847853A (en) * 2016-04-01 2016-08-10 深圳市创云方网络科技有限公司 Video content distribution method and device

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102511043A (en) * 2011-11-26 2012-06-20 华为技术有限公司 Method for replacing cache files, device and system thereof
CN103237068A (en) * 2013-04-17 2013-08-07 北京科技大学 Content-attribute-distinguishable streaming media cache replacement method for CDN-P2P (content delivery network and peer-to-peer)
CN104469502A (en) * 2014-11-11 2015-03-25 中原工学院 Cache replacement policy based on video similarity in relevance video-on-demand system
CN104822068A (en) * 2015-04-29 2015-08-05 四达时代通讯网络技术有限公司 Streaming media proxy cache replacing method and device
CN104967861A (en) * 2015-05-27 2015-10-07 上海美琦浦悦通讯科技有限公司 CDN video buffer system and method

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
信息中心网络中内置缓存块调度策略的研究;彭昊;《北京邮电大学硕士学位论文》;20161220;全文 *
基于IPTV用户行为分析的流媒体分均匀分段策略研究;邓宇珊;《华南理工大学硕士学位论文》;20150423;全文 *
视频分发与缓存协同优化技术研究;姚士佳;《中国科学技术大学博士学位论文》;20140420;全文 *

Also Published As

Publication number Publication date
CN107911711A (en) 2018-04-13

Similar Documents

Publication Publication Date Title
CN107911711B (en) Edge cache replacement improvement method considering partitions
CN107889160B (en) Small cell network edge part caching method considering user time delay
CN108600998B (en) Cache optimization decision method for ultra-density cellular and D2D heterogeneous converged network
CN112218337A (en) Cache strategy decision method in mobile edge calculation
CN106998353B (en) Optimal caching configuration method for files in content-centric networking
CN108093435B (en) Cellular downlink network energy efficiency optimization system and method based on cached popular content
CN105979274A (en) Distributive cache storage method for dynamic self-adaptive video streaming media
CN108737507B (en) D2D wireless caching method
Hajiakhondi-Meybodi et al. Cache replacement schemes based on adaptive time window for video on demand services in femtocell networks
CN112637908B (en) Fine-grained layered edge caching method based on content popularity
US9521064B2 (en) Cooperative caching method and apparatus for mobile communication system
CN110913239B (en) Video cache updating method for refined mobile edge calculation
CN111698732A (en) Time delay oriented cooperative cache optimization method in micro-cellular wireless network
Huang et al. D2D-assisted VR video pre-caching strategy
Miao et al. Multi-level plru cache algorithm for content delivery networks
Wang et al. Segment-based joint caching and recommendation optimization for mobile video transmission
CN113965962A (en) Cache resource management method and system for Internet of things slices
CN112822275B (en) Lightweight caching strategy based on TOPSIS entropy weight method
CN108429919B (en) Caching and transmission optimization method of multi-rate video in wireless network
CN114245422A (en) Edge active caching method based on intelligent sharing in cluster
CN113489779A (en) Accurate cache placement method based on network topology layering
CN109982389B (en) Wireless caching method based on online learning of multi-target multi-arm gambling machine
CN113709853A (en) Network content transmission method and device oriented to cloud edge collaboration and storage medium
Zha Key technologies of cache and computing in 5G mobile communication network
Meint et al. From FIFO to predictive cache replacement

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant