WO2013075341A1

WO2013075341A1 - Cache file replacement method, apparatus and system

Info

Publication number: WO2013075341A1
Application number: PCT/CN2011/082996
Authority: WO
Inventors: 凌强; 徐理想; 陆海翔; 王凤; 陈骞
Original assignee: 华为技术有限公司
Priority date: 2011-11-26
Filing date: 2011-11-26
Publication date: 2013-05-30
Also published as: CN102511043A; CN102511043B

Abstract

Disclosed are a cache file replacement method, apparatus and system. The method includes the following steps: acquiring a historical access frequency and a current access frequency of each stored cache file; obtaining an average access frequency of each cache file according to the historical access frequency, current access frequency, a preset historical access frequency weight value, a preset current access frequency weight value of each cache file, wherein the preset historical access frequency weight value is less than the preset current access frequency weight value; and acquiring a cache file with the smallest average access frequency and replacing the cache file with the smallest average access frequency with a new cache file. The cache file replacement method provided by the present invention can improve the effectiveness of cache policies. In addition, also provided are a cache file replacement apparatus, and a network system using the cache file replacement method.

Description

Cache file replacement method, device and system

The present invention relates to the field of computer communications, and in particular, to a cache file replacement method, apparatus, and system. Background technique

With the popularity of the network, the data transmitted by the network is not limited to text, but also includes video data. Due to the large code rate and long-term transmission of video VOD (Video On Demand) applications, the bandwidth and response speed of VOD servers are very high.

Therefore, it is considered to use a limited storage space to store part of the video content as a local proxy server to provide data to the user, thereby reducing the data traffic of the network trunk. As shown in Figure 1, 1, 2, 3 are streaming media file servers, 4 is a proxy server, and user 5 enjoys streaming media services through a proxy server.

Since the cache space of the proxy server is limited, when this space is full, some low-utilization data must be removed to provide better services. The cache replacement strategy is used to decide which cache data to remove, and its goal is to make better use of the available space resources.

There are two typical replacement strategies at present:

1. LRU (Least Recently Used) algorithm, which maintains a cache entry queue in which the cache entries are sorted by the last accessed time of each entry. When the cache space is full, the cache entry at the end of the queue (that is, the one that is the oldest at the time of the last access) is deleted, and the new zone is entered into the queue.

2. LFU (Least Frequently Used) algorithm, which sorts the blocks in the cache according to the frequency of access of each cache block. When the cache space is full, replace the access frequency in the cache queue. One.

However, both of the above algorithms have many problems. The LRU algorithm is more affected by volatility. For example, if a movie is accidentally accessed once and is no longer accessed, the movie will be put into the cache when the movie is processed, and a more popular movie may be deleted from the cache.

The LFU algorithm only maintains the accessed frequency information of each item, and for a cache entry that has a very high access frequency in the past and has a lower frequency of recent access, when the cache space is full, the cache entry 4 is difficult to be cached. Replaced, which in turn leads to a drop in hit rate. Summary of the invention

The technical problem to be solved by the embodiments of the present invention is to provide a method, a device, and a system for replacing a cache file, which are used to solve the problem that the cache file with low access frequency cannot be replaced from the cache in time in the prior art, thereby causing access. The problem of falling hit rate.

An embodiment of the present invention provides a cache file replacement method, where the method includes the following steps: acquiring a historical access frequency and a current access frequency of each cache file that has been stored;

Obtain an average access frequency of each cache file according to the historical access frequency of the cache file, the current access frequency, a weight value of a preset historical access frequency, and a weight value of a preset current access frequency, where The weight value of the preset historical access frequency is smaller than the weight value of the preset current access frequency;

Obtain the cache file with the lowest average access frequency, and replace the cache file with the smallest average access frequency with the new cache file.

Correspondingly, the embodiment of the present invention further provides a cache file replacement device, where the device includes: an obtaining module, configured to acquire a historical access frequency and a current access frequency of each cache file that has been stored;

The access frequency module is configured to obtain an average of each cache file according to the historical access frequency of the cache file, the current access frequency, a weight value of a preset historical access frequency, and a weight value of a preset current access frequency. An access frequency, where a weight value of the preset historical access frequency is less than a weight value of the preset current access frequency;

The replacement module is configured to obtain a cache file with the smallest average access frequency, and replace the cache file with the smallest average access frequency with the new cache file.

A network system, the network system includes a streaming media file server and a proxy server, wherein the proxy server includes the cache file replacement device, obtains a streaming media file from the streaming media file server, and the streaming media The file is saved as a cache file within the proxy server.

The cache file replacement method provided by the embodiment of the present invention can not only maintain a certain stability according to the old average access time interval, but also reduce the influence of the volatility of the latest access information. On the other hand, by occupying the proportion of the old average access time interval in the new average access time interval to be smaller than the current access interval, it is possible to gradually weaken the relatively long-term access information to the current access information after a certain period of time. The effect of avoiding the long-term access information interferes with the current sorting after the access mode changes, thus accurately reflecting the current access status of the cache file. In addition, the cache file replacement method provided by the embodiment of the present invention further predicts the average access time interval of the next moment by using the prediction formula, so that the cache file is accessed the latest time, and the closer the current time is, the greater the heat is obtained. The effect of the recent access state of the cache file on the heat. The heat is made closer to the current real access state of the cache file.

The cache file replacement method provided by the embodiment of the present invention further determines the actual relative size of the heat of two video segments by adding a weighting factor in the value formula. BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings, which are incorporated in the claims Other drawings may also be obtained from these drawings without the inventive labor.

1 is a schematic diagram of an existing streaming media network;

2 is a flowchart of a method for replacing a cache file according to the first embodiment of the present invention;

3 is a schematic diagram of a cache file replacing apparatus according to a first embodiment of the present invention;

4 is a flowchart of a method for replacing a cache file according to a second embodiment of the present invention;

FIG. 5 is a schematic diagram of a cache file replacing apparatus according to a second embodiment of the present invention; FIG.

6 is a flowchart of a method for replacing a cache file according to a third embodiment of the present invention;

FIG. 7 is a flowchart of a cache file replacing apparatus according to a third embodiment of the present invention. detailed description

BRIEF DESCRIPTION OF THE DRAWINGS The technical solutions in the embodiments of the present invention will be described in detail with reference to the accompanying drawings. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present invention without creative work are within the scope of the present invention.

The cache file replacement method provided by the embodiment of the present invention can reduce the influence of the volatility of the latest access information, and avoids that the access information long ago interferes with the current order after the access mode changes.

Referring to FIG. 2, a cache file replacement method according to a first embodiment of the present invention is provided. this method Used for the replacement of individual cache files in the proxy server. The method includes the following steps:

Step 101: Obtain a historical access frequency and a current access frequency of each cache file that has been stored. The historical access frequency may be expressed in terms of the number of visits in a past period of time, or may be expressed in an average time interval of visits over a period of time. The current access frequency may be represented by the number of accesses for the most recent period of time from the current time, or by the time interval of access during the period of time closest to the current time. In this embodiment, the historical access frequency is represented by an average access time interval in which the cache file is accessed, and the current access frequency is used by the time when the cache file is currently accessed and the cache file is last used. The time interval between the moments of access is indicated. In this step, the time when the cache file is currently accessed, the time when the cache file was last accessed, and the average access time interval in which the cache file is accessed are obtained. The cache file can be a video clip in a video or a complete video. In this embodiment, the cache file is a video segment in a video. The proxy server can divide a large video file into video segments of equal length according to the preset time period, such as dividing the 20-minute video into 0-5 minutes, 5-10 minutes, 10-15 minutes, 15- Four video clips in 20 minutes. When the user accesses the video, the proxy server first converts the user's request for the video file into a request for the corresponding video segment of the video file, such as a video clip corresponding to 0-5 minutes at the beginning. In this step, when a certain cache file is accessed by the user, the time when the cache file is currently accessed, the time when the cache file saved in the proxy server is last accessed, and the average access of the cache file are accessed. time interval. The initial value of the average access time interval is set to the current accessed time minus the time at which the cache file was last accessed.

Step 103: Obtain an average access frequency of each cache file according to the historical access frequency of the cache file, the current access frequency, a weight value of a preset historical access frequency, and a weight value of a preset current access frequency. The weight value of the preset historical access frequency is smaller than the weight value of the preset current access frequency.

In this embodiment, a preset time interval formula and a value formula are used to obtain an average access frequency of each cache file. The time interval formula and value formula are:

Γ = (1-α) χΓ ₀ + αχ (^2-Μ); (1)

M =丄; (2)

T

Where is the updated average access interval; "is a forgetting factor and satisfies 0.5<"<1; T ₀ is the average access time interval in which the cache file is accessed; N1 is the time at which the cache file is currently accessed; The time when the cache file was last accessed; Μ is the heat of the cache file, that is, the Average access frequency. The "weight value for the preset current access frequency, [alpha]-^ is the weight value of the preset historical access frequency.

In the above time interval formula, r. It is the old average access time interval determined by the previous access situation, and the value of the value can more accurately reflect the real access status of the cache file in the past period of time. Said r. The initial value is set to the time of the current access minus the time when the cache file was last accessed. The old average access interval r. After the latest access information (the time N2 at which the cache file is currently accessed) is combined by the forgetting factor _α , the influence of the volatility of the latest access information is reduced, thereby achieving the object of the present invention. On the other hand, it is a forgetting factor and takes values between 0.5 and 1. In the present embodiment, it can be assumed that the α = 0.7. Of course, in other embodiments, the values may also take other values. In the new average access interval, the old average access interval is Γ. The proportion is less than the current access interval (N2 - N1). So every time the film is visited, the previous visit is awkward. In the proportion of Γ, it will be reduced by a certain ratio. After a certain period of time, the proportion of the visit information that is relatively old now will become small until it is negligible. This avoids the long-standing access information from interfering with the current ordering after the access mode changes, thereby achieving the purpose of the time-varying characteristics of the present invention. This avoids the impact of cache files that have a very high frequency of access in the past and have recently been accessed less frequently.

Step 105: Obtain a cache file with the smallest average access frequency, and replace the cache file with the smallest average access frequency with the new cache file.

The larger the value calculated by the heat, the higher the heat. That is, the more frequently the cache file is accessed recently.

Sorting the plurality of cache files according to the size of the cache file. By sorting the heat of the cache file, it is possible to obtain which cache files are frequently accessed recently, and which files are recently accessed less frequently, thereby providing a reference for the proxy server to replace the file.

Delete the cache file with the least heat, that is, delete the cache file with the longest average access interval (the average access frequency is the smallest), and store the new file obtained from the streaming file server in the proxy server to become the new one. Cache files to improve the effectiveness of cache decisions.

Please refer to FIG. 3 , which is a cache file replacing apparatus 200 according to a first embodiment of the present invention. The cache file replacement device 200 is disposed in a proxy server 110. The proxy server 110 and the streaming media file server 120 form a network system 100. The proxy server 110 saves a part of the streaming media files in the streaming media file server 120 accessed by the user as a cache file. The hard disk (not shown) of the proxy server 110 is described. The cache file replacement device 200 includes an acquisition module 210, an access frequency module 220, and a replacement module 230.

The obtaining module 210 is configured to obtain a historical access frequency and a current access frequency of each cache file that has been stored. In this embodiment, the obtaining module 210 is configured to acquire a time when the cache file is currently accessed, a time when the cache file was last accessed, and an average access time interval in which the cache file is accessed. In this embodiment, when a cache file is accessed by the user, the time when the cache file is currently accessed, the time when the cache file saved in the proxy server is last accessed, and the average value of the cache file are accessed are obtained. Access time interval.

The access frequency module 220 is configured to obtain each cache file according to the historical access frequency of the respective cache file, the current access frequency, a weight value of a preset historical access frequency, and a weight value of a preset current access frequency. The average access frequency, wherein the weight value of the preset historical access frequency is smaller than the weight value of the preset current access frequency. In this embodiment, the access frequency module 220 is configured to: according to the obtained time when the cache file is currently accessed, the time when the cache file was last accessed, the average access time interval of the cache file being accessed, and the pre-emption The time interval formula is updated to update the average access time interval in which the cache file is accessed, and the time interval formula is: r = (l - a) xr ₀ + ax (N2 - M); where T is the updated average Access interval; a is a forgetting factor, and satisfies 0.5<a<l; T ₀ is the average access time interval in which the cache file is accessed; N2 is the time at which the cache file is currently accessed; N1 is on the cache file. The time of being visited. The access frequency module 220 calculates the average access time interval in which the cache file is accessed by using the method in step 103.

The access frequency module 220 includes a thermal sub-module 221 . The heat sub-module 221 is configured to calculate the heat of the cache file according to the updated average access time interval and a value formula, where the value formula is:

M= .

The replacement module 230 is configured to replace the cache file with the smallest average access frequency. In this embodiment, the replacement module 230 includes a sorting sub-module 231.

The sorting sub-module 231 is configured to pair the plurality of caches according to the size of the heat of the cache file. The files are sorted. Sorting the heat of the cache file enables the most frequently accessed cache files to be easily deleted in the cache space, and the cache files with lower recent access frequencies can be deleted faster in the cache space. The replacement module 230 is configured to replace the new file obtained from the streaming media file server with the cache file with the least heat.

Referring to FIG. 4, a cache file replacement method according to a second embodiment of the present invention is provided. The method includes the following steps:

Step 301: Obtain the historical access frequency, the current access frequency, and the access frequency of the next time of each cache file.

In this embodiment, the time at which the cache file is currently accessed, the time when the cache file was last accessed, and the average access time interval in which the cache file is accessed are obtained. In this embodiment, the above parameters are obtained by the same method as step 101. The access frequency of the next time instant may be represented by a time interval between the time at which the obtained cache file is currently accessed and the next time of the current time.

Step 303: Obtain an average access frequency of each cache file according to the historical access frequency of the cache file, the current access frequency, a weight value of a preset historical access frequency, and a weight value of a preset current access frequency. The weight value of the preset historical access frequency is smaller than the weight value of the preset current access frequency.

Γ = (1-α) χΓ ₀ + αχ (^2-Μ); (1) τ · ( 2 )

Step 305: Obtain each cache according to the current access frequency of each cache file, the access frequency of the next moment, the weight value of the preset current access frequency, and the weight value of the access frequency of the preset next time. The predicted access frequency of the file. In this embodiment, the average access frequency of each cache file is obtained according to formula (1) and the following two formulas,

T_&st = (l - c) x T'+c x (N _ est - N'); ( 3 )

=丄

T_est . ( 4 )

Wherein, the predicted average access time interval is the latest average access time interval, when the current time of the cache file is not accessed by the user, T' = T ₀ , when the current time of the cache file is accessed by the user , T' = T; c is a predictor, and satisfies 0.5<c<l; is the next moment of the current time; is the time when the cache file is last accessed, when the current time of the cache file is not used by the user When accessing, N' = N1, when the cache file is currently accessed by the user, N' = N ² . The c is a weight value of the access frequency of the preset next time.

When the cache file is not accessed by the user at the current time, Γ 'is the average access time interval of the cache file in the past, that is, Γ. . N is the time when the cache file was last accessed, that is, Μ. When the current time of the cache file is accessed by the user, N is the time at which the cache file is currently accessed, that is, N2.

The predictor has the same effect as the forgetting factor _Ω . In the present embodiment, the estimation factor c = 0.7 can be assumed. The next time N _ 6 ^ of the current time is 1 second added at the current time. Of course, in other embodiments, the predictor may take other values or take the same value as the forgetting factor _Ω . The next moment of the current moment may also be increased by 1 millisecond, 5 seconds or 1 minute at the current moment.

According to the actual user access situation, the historical average access time interval and the predicted next access time interval are combined with different weights to obtain an estimated value of the future access time interval.

Step 307: Replace the cache file with the least frequent access frequency at the next moment with the new cache file.

It can be seen from the prediction formula that the time N' at which the cache file is last accessed is closer to the cache file at the current time, and the shorter the interval time, the greater the heat M. Thereby, the influence of the most recent access state of the cache file on the heat M can be increased. The heat M is made closer to the current real access state of the cache file.

Sorting the plurality of cache files according to the size of the cache file. In this embodiment, the same method as step 105 is used to perform sorting, and the cache file with the least heat is replaced. Referring to FIG. 5, a cache file replacing apparatus 510 according to a second embodiment of the present invention is provided. The cache file replacement device 510 is substantially the same as the cache file replacement device 200 provided by the first embodiment, and is used in the same proxy server and network system. The cache file replacement device 510 includes an acquisition module 511, an access frequency module 512, a prediction module 513, and a replacement module 514.

The obtaining module 511 is configured to obtain a historical access frequency of each cache file, a current access frequency, and an access frequency of the next moment. In this embodiment, each parameter is obtained by the method of step 301.

The access frequency module 512 is configured to obtain each cache file according to the historical access frequency of the respective cache file, the current access frequency, a weight value of a preset historical access frequency, and a weight value of a preset current access frequency. The average access frequency, wherein the weight value of the preset historical access frequency is smaller than the weight value of the preset current access frequency. In this embodiment, a preset time interval formula and a value formula are used to obtain an average access frequency of each cache file. The time interval formula and value formula are:

Γ = (1-α) χΓ ₀ + αχ (^2-Μ); (1)

τ · ( 2 )

Where is the updated average access interval; "is a forgetting factor and satisfies 0.5<"<1; T ₀ is the average access time interval in which the cache file is accessed; N1 is the time at which the cache file is currently accessed; The time when the cache file was last accessed; Μ is the heat of the cache file, that is, the average access frequency. The weight value of the preset current access frequency, α-^ is the weight value of the preset historical access frequency. In this embodiment, the access frequency module 512 obtains each cache by the same method as step 303. The average frequency of access to the file.

The prediction module 513 is configured to use, according to the current access frequency of each cache file, the access frequency of the next moment, the weight value of the preset current access frequency, and the weight of the preset access frequency of the next moment. The value gets the predicted access frequency for each cache file. In this embodiment, the prediction module 513 calculates the average access time interval at the next moment in the same manner as the step 305. The prediction module 513 includes a thermal sub-module 513a for calculating the popularity of the cache file according to the average access time interval calculated by the prediction formula and the formula (4).

The replacement module 514 is configured to replace the cache file with the least frequent access frequency at the next moment with a new cache file. In this embodiment, the replacement module 514 includes a sorting sub-module 514a.

The sorting sub-module 514a is configured to sort a plurality of the cache files according to the size of the hotness of the cache file. In this embodiment, the ordering submodule 514a is the same as in the first embodiment. The sorting sub-module 231 is the same. The replacement module 514 is configured to replace the new file obtained from the streaming media file server with the cache file with the least heat.

Please refer to FIG. 6, which is a cache file replacement method according to a third embodiment of the present invention. In this embodiment, the cache file processed by the cache file replacement method is a video clip. The method includes the following steps:

Step 601: Acquire multiple videos, divide each video into at least one video segment, and use the same label rule label for each video segment in different videos.

In the present embodiment, each video is divided into a plurality of segments in a period of 5 minutes, and each video segment in each video is sequentially labeled with 0, 1, 2... in chronological order. Of course, other numbers or letter numbers may be used in other embodiments.

Step 603: Obtain a cache file historical access frequency and a current access frequency and a label of the video segment.

In this embodiment, the time when the video clip is currently accessed, the time when the video clip was last accessed, the average access time interval in which the video clip is accessed, and the label of the video clip are obtained. In this embodiment, the current time interval of the video segment is obtained by the same method as that in step 101. In addition, in this step, the label of each video segment is also obtained.

Step 605: Obtain an average access frequency of each cache file according to the historical access frequency of the cache file, the current access frequency, a weight value of a preset historical access frequency, and a weight value of a preset current access frequency. The weight value of the preset historical access frequency is smaller than the weight value of the preset current access frequency.

Γ = (1-α) χΓ ₀ + αχ (^2-Μ); (1) τ · ( 2 )

Where is the updated average access interval; "is a forgetting factor and satisfies 0.5<"<1; T ₀ is the average access time interval in which the cache file is accessed; N1 is the time at which the cache file is currently accessed; The time when the cache file was last accessed; Μ is the heat of the cache file, that is, the average access frequency. The "weight value for the preset current access frequency, α-^ is a preset historical visit Ask the weight value of the frequency.

Step 607: Obtain each cache according to the current access frequency of each cache file, the access frequency of the next moment, the weight value of the preset current access frequency, and the weight value of the access frequency of the preset next time. The predicted access frequency of the file.

In this embodiment, the average access frequency of each cache file is obtained according to formula (1) and the following two formulas,

T_&st = (l-c)xT'+cx(N_est-N'); (3)

Wherein, the predicted average access time interval is the latest average access time interval, when the current time of the cache file is not accessed by the user, T' = T ₀ , when the current time of the cache file is accessed by the user , T′ = T ; c is a predictor, and satisfies 0.5<c<l; is the next moment of the current time; is the time when the cache file is last accessed, when the current time of the cache file is not used by the user When accessing, Ν' ₌ Ν\, when the cache file is currently accessed by the user, N' = N1. The c is a weight value of the access frequency of the preset next time.

Step 609: Obtain an access probability of the cache file.

In this embodiment, an average access probability of each video segment is calculated according to a preset access probability formula, and the access phase difference rate formula is:

Tbar[j] = a2 Tbar[j] + (1 - a∑) x X _? J =0123 . (5 )

Where j is the label of the video segment; 73⁄4ar[ | is the average on-demand probability of the video segment labeled j, the initial value is 1; 2 is a constant, 2 is greater than or equal to 0.9, and is less than 1; X is an access State coefficient, when the video segment is accessed, the X = l, otherwise Χ = 0.

In the present embodiment, the j = 0, 1, 2, 3, .... 73⁄4r[ | The average on-demand probability preset for the jth video clip. 2 is 0.999. Of course, in other embodiments, the 2 may also be other fractions that are smaller than and close to 1. For example, if the 0th video clip is accessed for the first time, then 3⁄4ar[0]= 0.999 X 1+(1-0.999) x l=l. The remaining video clips 3⁄4ar[l], Tbar[2]. 3⁄4 [3] are equal to =0.999 x 1+(1-0.999) x

0=0.999. The above formula is equivalent to a low-pass filter, and 2 = 0.999 can make the filter bandwidth small, thereby reducing the randomness of user access behavior.

Calculating the calculated average access probability of each of the video segments according to a preset normalization formula, the normalization formula is:

v ₌ TbaAm\ . (6)

Tbar[j]

Where m is any number in j; P_k[j] is the normalized average access probability of the video segment labeled j. In the present embodiment, all 3⁄4arj] are normalized based on a 3⁄4ar[m]. Said

P_k[j] is an array of intermediate variables.

The weighting parameter is calculated according to the normalized average access probability and the parameter formula, and the parameter formula is:

P[j] = a3xP[j] + (l-a3)xy; (7)

Where P[j] is the weighting parameter of the video segment labeled j, and its initial value is 1; α3 is a constant, greater than or equal to 0.9, and less than 1; if P_kj] >1000, Bay ¹ Jy = 1000, if P_k[ j] ≤ 1000 , Bay ¹ Jy = P_k[j].

The above formula is also a low-pass filter in order not to make P[j] change too drastically. Limit the maximum value of y to no more than 1000, avoiding the perturbations caused by too severe weighting.

All of the calculated weighting parameters are adjusted to be equal to the weighting factor b[j] of T_est in the same ratio. In order not to change the relative heat size of the video segment after weighting too much, it is necessary to reduce the weighting parameter corresponding to different video segments to a number smaller than D_681. In this embodiment, when the experiment finds that the power of the quarter of the weighting parameter is taken, the hit rate of the video segment on the proxy server is high. Therefore, the adjustment formula b[ ] = ρϋ] is adopted. • ²⁵ adjusts all of the weighting parameters to the same scale to a number less than T_est. Of course, in other embodiments, the force weight parameter may also be adjusted by multiplying by a decimal or by a percentage.

Step 611: Replace the cache file with the lowest access frequency and the lowest access probability at the next moment with the new cache file.

In this embodiment, the heat of the cache file is calculated according to the weighting factor and a value formula, and the value formula is:

M = ^ ¹ - ^; (8)

T_estxb[j]

Due to the influence of volatility, the heat M size of two video clips at a certain time may be similar. In this case, it is not easy to determine the true relative size of the heat scores of the two video clips. In this embodiment, the importance of the video segment (i.e., the weighting factor b[j']) is obtained by statistical analysis of the historical access data of the video segment. The heat degree M is evaluated by combining the weighting factor b[j'] with the average access time interval T_est of the next moment of the video clip. Thus using trend information in historical access information to the value function

M =—— ¹ -——, filtering out the influence of current volatility, thus determining the heat of two video clips M T_estxb[j]

The actual relative size, and replaces the cache file with the least frequent access frequency and the lowest probability of access at the next moment with the acquired new file.

In this embodiment, when the video segment is accessed, the current access situation and the historical access situation are combined by the formula TbarU] = 2x Tbar[j] + (1-α2) , to obtain an average on-demand probability of the video segment; Then, the obtained average on-demand probability 73⁄4ar[ ] is normalized by the set method to obtain an intermediate parameter P_k[j]; the intermediate parameter p_kj] obtained this time and the historical intermediate parameter P j] are given by the formula P. [j]=a3 xP[j]+(l-a3)xy are combined to obtain an average intermediate parameter P[j]; Finally, the resulting flat The intermediate intermediate parameter P j] uses the formula b '] = p[j]. ²⁵ to adjust, get a new, reasonable weighting factor bj]. This method not only ensures that the weighting factor bj] can reflect the latest user access situation in real time, but also avoids excessive changes in the weighting factor b[j] to cause disturbance to the decision.

Referring to FIG. 7, a cache file replacing apparatus 700 according to a third embodiment of the present invention is provided. The cache file replacement device 700 includes a segmentation module 710, an acquisition module 720, an access frequency module 730, a prediction module 740, a probability module 750, and a replacement module 760.

The segmentation module 710 is configured to acquire a plurality of videos, and divide each video into at least one video segment, and the step 601 performs segmentation and labeling in the same manner.

The obtaining module 720 is configured to obtain a cache file historical access frequency and a current access frequency and a label of the video segment. In this embodiment, the time when the video clip is currently accessed, the time when the video clip was last accessed, the average access time interval in which the video clip is accessed, and the video clip are obtained by using the same method in step 603. The label.

The access frequency module 730 is configured to obtain each cache file according to the historical access frequency of the respective cache file, the current access frequency, a weight value of a preset historical access frequency, and a weight value of a preset current access frequency. The average access frequency, wherein the weight value of the preset historical access frequency is smaller than the weight value of the preset current access frequency. In this embodiment, the average access time interval is calculated by the same method as step 605.

The prediction module 740 is configured to use, according to the current access frequency of each cache file, the access frequency of the next moment, the weight value of the preset current access frequency, and the weight of the preset access frequency of the next moment. The value gets the predicted access frequency for each cache file. In the present embodiment, the predicted access frequency of each cache file is calculated in the same manner as the step 607.

The probability module 750 is configured to obtain an access probability of the cache file. In this embodiment, the probability module 750 includes a probability sub-module 751, a normalization sub-module 752, a weighting parameter sub-module 753, and an adjustment sub-module 754.

The probability sub-module 751 is configured to calculate an average access probability of each video segment according to a preset access probability formula, where the access probability formula is:

Tbar[j] = a2 Tbar[j] + (1 - a∑) x X _? J =0 1 2 3 . ( 5 )

Where j is the label of the video segment; 73⁄4ar[ | is the average on-demand probability of the video segment labeled j, the initial value is 1; 2 is a constant, 2 is greater than or equal to 0.9, and is less than 1; X is an access State coefficient, when the video segment is accessed, the x = l, otherwise x = 0.

In the present embodiment, the j = 0, 1, 2, 3, .... Rbar[ ] is the average on-demand probability preset for the jth video segment. 2 is 0.999. Of course, in other embodiments, the 2 may also be other fractions that are smaller than and close to 1. For example, if the 0th video clip is accessed for the first time, then 3⁄4ar[0]= 0.999 X 1+(1-0.999) x l=l. The remaining video clips 3⁄4ar[l], Tbar[2]. 3⁄4 [3] are equal to =0.999 x 1+(1-0.999) x 0=0.999. The above formula is equivalent to a low-pass filter, and 2 = 0.999 can make the filter bandwidth small, thereby reducing the randomness of user access behavior.

The normalization sub-module 752 is configured to normalize the calculated average access probability of each of the video segments according to a preset normalization formula, where the normalization formula is:

p _{kU] =} 73⁄4orM . ( ₆ )

Tbar[j]

Where j=0, l, 2, 3...; m is any number in j; p_kj] is the normalized average access probability of the video segment labeled j. In the present embodiment, all Tbarlj] are normalized based on a 3⁄4ar[m]. The p_kj] is an array of intermediate variables.

The weighting parameter sub-module 753 is configured to calculate a force port weight parameter according to a normalized average access probability and a parameter formula, wherein the parameter formula is: P j] = 3xP[j] + (l-3)xy; )

Where P[j] is the weighting parameter of the video segment labeled j, and its initial value is 1; α3 is a constant, 3 is greater than or equal to 0.9, and is less than 1; #P_k j] >1000, Bay ¹ J y = 1000, If P_k[j] ≤ 1000, Bay ¹ J y = P_k[j]. The above formula is also a low-pass filter in order not to make P[j] change too drastically. Limit the maximum value of y to no more than 1000, avoiding the disturbance caused by too severe weighting.

The adjustment sub-module 754 is configured to adjust all the calculated weighting parameters to a weighting factor b[j] smaller than T_est in the same ratio. In the present embodiment, all the weighting parameters are adjusted to the weighting factor smaller than T_est by the same ratio using the adjustment formula b[j]=p[j] ²⁵ .

The weighting factor bj] may also be set according to an empirical value, thereby omitting the normalization sub-module 752, the weighting parameter sub-module 753, and the adjustment sub-module 754. Other formulas that can implement the functions of the normalization sub-module 752, the weighting parameter sub-module 753, and the adjustment sub-module 754 can also be used in the weighting factor bj].

The replacement module 760 also includes a thermal sub-module 761 and a sequencing sub-module 762.

The heat sub-module 761 is configured to calculate the heat of the cache file according to the weighting factor and a value formula, and the value formula is:

M = ^ ¹ - ^. (8)

T_estxb[j]

The sorting sub-module 762 is configured to sort a plurality of the cache files according to the size of the heat of the cache file. The replacement module 760 is for a new file to be obtained from a streaming file server Replace the least hot cache file.

The cache file replacement method and apparatus provided by the embodiments of the present invention can not only maintain a certain stability according to the old average access time interval, but also reduce the influence of the volatility of the latest access information. On the other hand, by occupying the proportion of the old average access time interval in the new average access time interval to be smaller than the current access interval, it is possible to gradually weaken the relatively long-term access information to the current access information after a certain period of time. The effect of avoiding the long-term access information interferes with the current sorting after the access mode changes, thus accurately reflecting the current access status of the cache file.

In addition, the cache file replacement method and apparatus provided by the embodiment of the present invention further predicts the average access time interval of the next moment by using the prediction formula, so that the cache file is accessed the latest time, and the closer to the current time, the greater the heat is obtained. The effect of the recent access state of the cache file on the heat. The heat is made closer to the current real access state of the cache file.

The cache file replacement method and apparatus provided by the embodiment of the present invention further increases the weight of the two video segments by statistically analyzing the historical access data by adding a weighting factor to the value formula to determine the actual heat of the two video segments. Relative size.

In the several embodiments provided herein, it should be understood that the disclosed systems, devices, and methods may be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the unit is only a logical function division. In actual implementation, there may be another division manner, for example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored, or not executed. In addition, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be electrical, mechanical or otherwise.

The units described as separate components may or may not be physically separate, and the components displayed as the units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solution of the embodiment.

In addition, each functional unit in each embodiment of the present invention may be integrated into one adjustment unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional units.

The above integrated unit implemented in the form of a software functional unit can be stored in a computer Read in the storage medium. The software functional unit described above is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, server, or network device, etc.) to perform portions of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: a USB flash drive, a mobile hard disk, a read-only memory (Read-Only Memory), a random access memory (RAM), a disk or an optical disk, and the like. The medium of the program code.

It should be noted that the above embodiments are only for explaining the technical solutions of the present invention, and are not intended to be limiting; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those skilled in the art that: The technical solutions described in the foregoing embodiments are modified, or some of the technical features are equivalently replaced. The modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

Rights request

A cache file replacement method, the method comprising the following steps:

Obtain the historical access frequency and current access frequency of each cache file that has been stored;

The cache file replacement method according to claim 1, wherein the obtaining the stored historical access frequency and the current access frequency of each cache file further comprises: acquiring an access frequency of the next time;

And after the step of obtaining the average access frequency of each cache file according to the historical access frequency of each cache file, the current access frequency, the weight value of the preset historical access frequency, and the weight value of the preset current access frequency, :

Obtaining predictions of each cache file according to the current access frequency of each cache file, the access frequency of the next moment, the weight value of the preset current access frequency, and the weight value of the access frequency of the preset next time An access frequency, wherein a weight value of the preset current access frequency is smaller than a weight value of an access frequency of the preset next time;

Correspondingly, the obtaining the cache file with the smallest average access frequency and replacing the cache file with the smallest average access frequency with the new cache file further includes: replacing the cache file with the least frequent access frequency at the next moment with the new cache file. .

The cache file replacement method according to claim 2, wherein the weight value of the current access frequency, the access frequency of the next moment, and the preset current access frequency according to each cache file After the step of obtaining the predicted access frequency of each cache file by the preset weight value of the access frequency at the next moment, the method further includes:

Obtaining an access probability of each cache file;

Correspondingly, the replacing the cache file with the least frequent access frequency at the next moment by using the new cache file comprises: replacing, by the new cache file, the cache file with the lowest access frequency and the lowest access probability at the next moment.

The cache file replacement method according to any one of claims 1 to 2, wherein the historical access frequency, the current access frequency, and a preset according to the respective cache files are The weight value of the historical access frequency and the weight value of the preset current access frequency obtain the average access frequency of each cache file including:

Γ = (1-α)χΓ ₀ +ax(N2-M);

M =丄;

T

Wherein, is the updated average access interval; "is a forgetting factor, and satisfies 0.5<"<1; ^ is the average access time interval in which the cache file is accessed; N1 is the time at which the cache file is currently accessed; The time when the cache file was last accessed; M is the average access frequency.

The cache file replacement method according to claim 4, wherein the weight of the current access frequency, the access frequency of the next moment, and the preset current access frequency according to each cache file The value, the preset weight value of the access frequency of the next time to obtain the predicted access frequency of each cache file includes:

R_est = (l-c)xT'+cx(N _est-N');

^丄;

T_est

Wherein, the average access time interval for the future is predicted; for the latest average access time interval, when the current time of the cache file is not accessed by the user, T'= ^T when the current time of the cache file is accessed by the user, T ' = T ; c is the predictor, and satisfies 0.5 < c <l; is the next moment of the current time; is the time when the cache file is last accessed, when the current time of the cache file is not accessed by the user , N' = N1, when the cache file is currently accessed by the user, N' ₌ N1.

The cache file replacement method according to claim 5, wherein the cache file is a video segment, and the steps of the step of acquiring the historical access frequency and the current access frequency of each cache file that are stored include the following steps. :

Acquiring multiple videos, dividing each video into at least one video segment, and using the same labeling rule label for each video segment in the different video;

And acquiring, in the step of acquiring the historical access frequency and the current access frequency of the stored cache files, the label of the video segment;

The access frequency according to the current access frequency and the next moment according to each cache file In the step of obtaining a predicted access frequency of each cache file by using a weight, a weight value of the preset current access frequency, and a weight value of the access frequency of the preset next time, obtaining an average access frequency of each cache file according to the following formula;

Tbar[j] = a2x Tbar[j] + (1-α2)χΧ;

p _{kU] =} 73⁄4or[m] _{? j=0} ,l,2,3...;

Tbar[j]

P[j] = a3xP[j] + (l-a3)xy;

b[ ] = p[j] ⁰²⁵ ;

M ¹ -

T_est b[j]

Where j is the label of the video segment; ^^[ is the default on-demand probability of the video segment labeled j, the initial value is 1; " ² is a constant," ^{2 is} greater than or equal to 0.9, and is less than 1; X is An access state coefficient, when the video segment is accessed, the χ=ι, otherwise χ=ο; is any number in j; P- ¹ *'] is the normalized video segment labeled j Average access probability; ^ρ ϋ] is the weighting parameter of the video segment labeled j, whose initial value is 1; " ³ is a constant," ^{3 is} greater than or equal to 0.9, and is less than 1; if !^_]^]>1000, Then y = 1000, if !^_]^] ≤ 1000, then y = P_k[j] _; Wj'] is the weighting factor.

7. A cache file replacement device, the device comprising:

An obtaining module, configured to obtain a historical access frequency and a current access frequency of each cache file that has been stored;

The cache file replacement device according to claim 7, wherein the acquisition module is configured to acquire an access frequency of a next time, the cache file replacement device further comprising:

a prediction module, configured to obtain, according to the current access frequency of each cache file, the access frequency of the next moment, the weight value of the preset current access frequency, and the weight value of the access frequency of the preset next moment a predicted access frequency of each cache file, where the weight value of the preset current access frequency is smaller than a weight value of the access frequency of the preset next time;

Correspondingly, the replacing module is configured to replace the access of the next moment with a new cache file. The least frequently cached file.

The cache file replacement method according to claim 8, wherein the cache file replacement device further comprises:

a probability module, configured to acquire an access probability of the cache file;

Correspondingly, the replacement module is further configured to replace, by the new cache file, the cache file with the lowest access frequency and the lowest probability of access at the next moment.

The cache file replacement device according to any one of claims 7, 8 or 9, wherein the access frequency module is configured to obtain an average access frequency of each cache file according to the following formula:

Γ = (1-α)χΓ ₀ +αχ(^2-Μ);

M =丄;

T

Where r is the updated average access interval; _Ω is the forgetting factor and satisfies 0.5 < _Ω <1; r. The average access time interval for which the cache file is accessed; N2 is the time at which the cache file is currently accessed; Μ is the time when the cache file was last accessed; Μ is the average access frequency.

11. The cache file replacement device according to claim 10, wherein the access frequency module is configured to obtain an average access frequency of each cache file according to the following formula:

R_est = (l-c)xT'+cx(N _est-N');

M =丄;

T_est

Wherein, the average access time interval for the future is predicted; for the latest average access time interval, when the current time of the cache file is not accessed by the user, τΆ, when the current time of the cache file is accessed by the user, T' = T; c is a predictor, and satisfies 0.5<c<l; is the next moment of the current time; is the time when the cache file is last accessed, when the current time of the cache file is not accessed by the user, N ' ₌ N\ , N' ₌ N1 when the cache file is currently accessed by the user.

The cache file replacement device according to claim 11, wherein the cache file is a video clip, and the cache file replacement device further includes:

a segmentation module, configured to acquire a plurality of videos, divide each video into at least one video segment, and use the acquiring module to acquire a label of the video segment;

The access frequency module is configured to obtain an average access frequency of each cache file according to the following formula, Tbar[j] = a2x Tbar[j] + (1-α2)χΧ;

p _{kU] =} 73⁄4or[m] _{? j=0} ,l,2,3...;

Tbar[j]

P[j] = a3xP[j] + (l-a3)xy; T_estxb[j]

Where j is the label of the video segment; "'] is the preset average on-demand probability of the video segment labeled j, with an initial value of 1; " ² is a constant," ^{2 is} greater than or equal to 0.9, and is less than 1; An access state coefficient, when the video segment is accessed, the X=l, otherwise X=0; ^ is any number of 』'; P- ¹ *'] is the normalization of the video segment labeled j The average access probability after the transformation; ^ρ ϋ] is the weighting parameter of the video segment labeled j, whose initial value is 1; " ³ is a constant," ^{3 is} greater than or equal to 0.9, and is less than 1; if !^_]^]> 1000, then y = 1000, if !^_]^] ≤ 1000, then y = P_k[j] _; Wj'] is the weighting factor.

A network system, comprising: a streaming media file server and a proxy server, wherein the proxy server comprises the cache file replacing device according to any one of claims 7-9, And obtaining, by the streaming media file server, a streaming media file, and saving the streaming media file as a cache file in the proxy server.