CN104683318A - An edge streaming media server cache selection method and system - Google Patents

Abstract

The present invention relates to a buffer selection method of an edge streaming media server, including: aggregating multiple users into several user categories according to their respective preferences; counting the strength of each user category, and each user category's preference for movies; wherein, the user category The strength of is the sum of the strengths of each user in the user class, and the user strength is the different influence of the user on the provider's caching decision; according to the degree of preference of the movie by each user class and the strength of each user class, calculate The utility of the movie; the movie with a larger utility value is selected to be deployed in the cache space of the edge streaming media server.

Description

An edge streaming media server cache selection method and system

technical field

The invention relates to the field of network communication, in particular to a buffer selection method and system for an edge streaming media server.

Background technique

In order to solve the performance bottleneck of the C/S architecture streaming media system and reduce operating costs, Content Delivery Network (CDN) is widely used. CDN places a large number of edge cache servers at the edge of the network. The content they cache can directly serve the user's on-demand request, avoiding the data throughput between the user and the backbone network, and achieving the purpose of reducing data transmission delay, smoothing transmission fluctuations, and reducing backbone network traffic.

For services such as digital interactive TV, streaming servers are deployed in each service area at the edge of the network. The streaming server is located between the user and the CDN, and its functions include: acting as an agent for the user's on-demand request, obtaining video data from the CDN, streaming and processing it into a format supported by IP-QAM, and pushing it to the user. The distance between the streaming server and the user is usually very close. In order to take full advantage of this advantage, the streaming server often also has a proxy cache function to store the previously obtained CDN video files and directly serve the subsequent same on-demand requests, further improving the user experience. Quality of service, reducing the burden on the CDN.

Because video files are usually large in size, both the CDN edge cache server and the streaming server that supports the proxy cache function are facing pressure on storage capacity. Therefore, a crucial issue is how to reasonably select the cache deployment content of the cache server, and effectively utilize its limited storage space, so as to give full play to the role of the cache and improve the overall service performance.

Contents of the invention

The purpose of the present invention is to overcome the defect in the prior art that cache deployment content of a cache server cannot be reasonably selected, thereby providing a method and system for selecting an edge streaming media server cache.

In order to achieve the above object, the present invention provides a buffer selection method for an edge streaming media server, comprising:

Step 1), aggregate multiple users into several user categories according to their preferences;

Step 2), the strength of each user class obtained in step 1), and the preference of each user class for movies; wherein, the strength of the user class is the sum of the strengths of each user in the user class, and the user strength is the user's preference for the movie. The differential influence that providers' caching decisions have;

Step 3), calculate the utility of the movie according to the degree of preference of the movie by each user category and the strength of each user category;

Step 4), select a movie with a large utility value and deploy it in the cache space of the edge streaming server.

In the above technical solution, the step 1) includes:

Step 1-1), define the user's preference for the movie according to the user's viewing time and viewing times of the movie;

Step 1-2), adding tags to the movie according to the attributes of the movie, and dividing the movie into movie categories according to the tags;

Step 1-3), obtain the user's preference value of the movie category to which the movie belongs based on the user's preference for a certain movie;

Step 1-4), cluster the users according to the user's preferences for each video category, and obtain several user categories.

In the above technical solution, the step 2) includes:

Step 2-1), calculate the user's activity level according to the frequency of the user's on-demand operation and the user's viewing time;

Step 2-2), set the service level of the user;

Step 2-3), calculate user strength according to user activity and service level;

Step 2-4), calculate the user class strength from the user strength;

Step 2-5), in a certain user category, quantify the recent activity of the user;

Step 2-6), using the user activity degree obtained in step 2-5) as the weight, to measure the preference of this user category for a certain movie.

In the above technical solution, in the step 3), the utility of the movie is calculated in the following way: taking the strength of the user class as the weight, and making a weighted sum of the preference values of each user class for the movie.

The present invention also provides an edge streaming media server cache selection system, including: a user class aggregation module, a user class strength and user class preference generation module, a movie utility calculation module, and a deployment module; wherein,

The user class aggregation module aggregates multiple users into several user classes according to their respective preferences;

The intensity of each user category obtained by the user category aggregation module and the user category preference generation module statistics, and the preference of each user category to the film; wherein, the intensity of the user category is the sum of each user intensity in the user category, The user strength is the different influence of the user on the provider's caching decision;

The film utility calculation module calculates the utility of the film according to the degree of preference of the film by each user category and the strength of each user category;

The deployment module selects movies with larger utility values and deploys them in the cache space of the edge streaming media server.

The advantages of the present invention are:

The present invention utilizes the clustering and recommendation algorithm to more accurately obtain user preference in the local area, which is used as the basis for cache deployment, thereby increasing the accuracy of judgment.

Description of drawings

Fig. 1 is a flow chart of the edge streaming media server cache selection method of the present invention.

Detailed ways

The present invention will be further described now in conjunction with accompanying drawing.

In this application, the edge cache server and the streaming server in the CDN are collectively referred to as the edge streaming server.

With reference to Fig. 1, method of the present invention comprises:

Step 1), multiple users are aggregated into several user categories according to their respective preferences.

In order to realize differentiated services in this application, it is necessary to discover each user's preference from the user's on-demand history. However, the information contained in the on-demand behavior of a single user is limited, so different users should be aggregated into several user categories according to their preferences for each movie, and then use collaborative filtering technology to analyze the importance of each movie for each user category .

This step specifically includes:

Step 1-1), first, define user u’s preference for movie v as:

in, Nu _{and Tu} respectively represent the number of video requests and viewing time of user u for v within a period of time, and the total number of video requests and viewing time of user u.

User u's preference for video v Constitutes a two-dimensional vector of user and preference values The missing items (that is, user u has not watched a certain movie) are zeroed, or predicted by the SVD model.

Step 1-2), generate movie class. Select several movie attributes to add tags to the movie object. The optional tag attributes include theme, release time, related characters, etc. Classify videos that share tag values into one movie category, and the same movie can enter multiple movie categories. Accurate and diverse classification labels are the basis for user clustering.

Steps 1-3): From the user's preference for a certain movie, the user's preference value for the movie category to which the movie belongs is obtained.

The user's preference value for a certain movie category is the sum of preferences for all movies of that category.

Assume that user u's preference value for a certain label t is: Then the user's preference vector can be defined as $P_u=({\mathrm{\ρ}}_u^1,{\mathrm{\ρ}}_u^2,\&Center\; Dot;\&Center\; Dot;\&Center\; Dot;,{\mathrm{\ρ}}_u^t,\&Center\; Dot;\&Center\; Dot;\&Center\; Dot;).$

Steps 1-4), finally, according to the user's preference for each video category, use the K-Means algorithm to cluster the users to obtain several user categories.

Step 2), count the intensity of each user class, and the user class's preference for movies.

The varying influence that users have on a provider's caching decisions is called user strength. The basis for defining user strength is that the higher the service level and the more active the user is, the greater the strength will be. Each user class has different numbers and importances of users included in each user class, and also presents different importance, which is called the user class strength, which is the sum of the user strengths in the user class. Among them, the user's activity is measured by the frequency of the user's on-demand operation and the user's viewing time, which is the weighted sum of the two; the user's service level depends on the service provider's non-technical strategy, which is the allocation of service resources The important basis of , the more important users have higher user level values.

The preference of a user class for a certain movie is defined as the weighted sum of the preferences of users in the class for the movie, and the user activity is used as the weight.

This step specifically includes:

Step 2-1), calculating the activity level of the user.

The user's activity level can be measured by two factors: the frequency with which the user performs on-demand operations and the user's viewing time.

Among them, Nu , T _u , N _U , and T _U respectively represent the number of video requests and movie viewing time of user u within a period of time, and _the total number of video requests and movie viewing time of user set U, and the α factor belongs to the real number interval [0,1 ], used to adjust the relative weights of the two factors.

Step 2-2), set the service level of the user. The user service level depends on the non-technical strategy of the service provider and is an important basis for allocating service resources. Users can be easily divided into member users UM and ordinary users UN.

Step 2-3), calculating user intensity.

The strength of user u is defined as:

$\stackrel{<span\; class="notranslate">\; ~</span>}{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; u</span>}}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; \{</span>\begin{array}{cc}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&beta;</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; u</span>}}^{\mathrm{<span\; class="notranslate">\; u</span>}}<span\; class="notranslate">\; )</span>& \mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; N</span>}}\\ \mathrm{<span\; class="notranslate">\; \&gamma;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&beta;</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; u</span>}}^{\mathrm{<span\; class="notranslate">\; u</span>}}<span\; class="notranslate">\; )</span>& \mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}\end{array}\mathrm{<span\; class="notranslate">\; \&gamma;</span>}<span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; \&beta;</span>}<span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; 0</span>}$

Among them, γ is used to distinguish between ordinary users and member users, β itself has no special meaning, and β is added just to prevent this item from being zero.

User intensity can be normalized:

${\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; u</span>}}<span\; class="notranslate">\; =</span>\frac{\stackrel{<span\; class="notranslate">\; ~</span>}{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\mathrm{<span\; class="notranslate">\; u</span>}}}}{{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}^{\stackrel{<span\; class="notranslate">\; ~</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; u</span>}}\stackrel{<span\; class="notranslate">\; ~</span>}{{\mathrm{<span\; class="notranslate">\; i</span>}}_{\stackrel{<span\; class="notranslate">\; ~</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}}}}$

After simulation evaluation, when α involved in step 2-1) and β and γ in this step are set to 0.4, 1, and 128 respectively, the method of the present invention has a very high cache hit rate.

Step 2-4), calculate the user class strength.

The strength of the user class is the sum of the strengths of all users in the user class, that is, the strength of the user class c is defined as

Step 2-5), within each user category c, quantify the recent activity of the user:

${\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; u</span>}}^{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; N</span>}}_{\mathrm{<span\; class="notranslate">\; u</span>}}}{{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}^{\stackrel{<span\; class="notranslate">\; \&Center\; Dot;</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; c</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}}_{\stackrel{<span\; class="notranslate">\; \&Center\; Dot;</span>}{\mathrm{<span\; class="notranslate">\; u</span>}}}}<span\; class="notranslate">\; .</span>$

Among them, ü represents any user in user class c.

Step 2-6), using the user activity obtained in step 2-5) as the weight, measure the preference of user class c for movie v:

${\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}^{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; =</span>\stackrel{\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; c</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; u</span>}}^{\mathrm{<span\; class="notranslate">\; v</span>}}{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; u</span>}}^{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; .</span>$

Step 3), according to the degree of preference of each movie by each user category and the strength of each user category calculated in step 2), calculate the utility of the movie.

The utility of a movie is defined as the weighted sum of the preference values of each user class for the movie, with the strength of the user class as the weight. Its calculation formula is as follows:

${\mathrm{<span\; class="notranslate">\; \&psi;</span>}}_{\mathrm{<span\; class="notranslate">\; v</span>}}<span\; class="notranslate">\; =</span>\stackrel{\mathrm{<span\; class="notranslate">\; c</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}^{\mathrm{<span\; class="notranslate">\; v</span>}}$

Step 4), select the movie with larger utility value, and deploy it in the limited cache space.

The present invention also provides an edge streaming media server cache selection system corresponding to the aforementioned selection method, the system comprising: a user class aggregation module, a user class strength and user class preference generation module, a movie utility calculation module, and a deployment module; wherein,

The user class aggregation module aggregates multiple users into several user classes according to their respective preferences;

The intensity of each user category obtained by the user category aggregation module and the user category preference generation module statistics, and the preference of each user category to the film; wherein, the intensity of the user category is the sum of each user intensity in the user category, The user strength is the different influence of the user on the provider's caching decision;

The film utility calculation module calculates the utility of the film according to the degree of preference of the film by each user category and the strength of each user category;

The deployment module selects movies with larger utility values and deploys them in the cache space of the edge streaming media server.

The method and system of the present invention use clustering and recommendation algorithms to more accurately obtain user preferences in the region, and use them as cache deployment basis to increase the accuracy of judgment.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than limit them. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art should understand that modifications or equivalent replacements to the technical solutions of the present invention do not depart from the spirit and scope of the technical solutions of the present invention, and all of them should be included in the scope of the present invention. within the scope of the claims.

Claims (5)

1. An edge streaming media server cache selection method, comprising: Step 1), aggregate multiple users into several user categories according to their preferences; Step 2), the strength of each user class obtained in step 1), and the preference of each user class for movies; wherein, the strength of the user class is the sum of the strengths of each user in the user class, and the user strength is the user's preference for the movie. The differential influence that providers' caching decisions have; Step 3), calculate the utility of the movie according to the degree of preference of the movie by each user category and the strength of each user category; Step 4), select a movie with a large utility value and deploy it in the cache space of the edge streaming server. 2. The edge streaming media server cache selection method according to claim 1, wherein said step 1) comprises: Step 1-1), define the user's preference for the movie according to the user's viewing time and viewing times of the movie; Step 1-2), adding tags to the movie according to the attributes of the movie, and dividing the movie into movie categories according to the tags; Step 1-3), obtain the user's preference value of the movie category to which the movie belongs based on the user's preference for a certain movie; Step 1-4), cluster the users according to the user's preferences for each video category, and obtain several user categories. 3. The edge streaming media server cache selection method according to claim 1, wherein the step 2) comprises: Step 2-1), calculate the user's activity level according to the frequency of the user's on-demand operation and the user's viewing time; Step 2-2), set the service level of the user; Step 2-3), calculate user strength according to user activity and service level; Step 2-4), calculate the user class strength from the user strength; Step 2-5), in a certain user category, quantify the recent activity of the user; Step 2-6), using the user activity degree obtained in step 2-5) as the weight, to measure the preference of this user category for a certain movie. 4. The edge streaming media server cache selection method according to claim 1, characterized in that, in the step 3), the utility of the movie is calculated in the following way: with the strength of the user class as the weight, for each user class A weighted sum is performed on the preference values of the movies. 5. An edge streaming media server cache selection system, comprising: a user class aggregation module, a user class strength and user class preference generation module, a movie utility calculation module and a deployment module; wherein, The user class aggregation module aggregates multiple users into several user classes according to their respective preferences; The intensity of each user category obtained by the user category aggregation module and the user category preference generation module statistics, and the preference of each user category to the film; wherein, the intensity of the user category is the sum of each user intensity in the user category, The user strength is the different influence of the user on the provider's caching decision; The film utility calculation module calculates the utility of the film according to the degree of preference of the film by each user category and the strength of each user category; The deployment module selects movies with larger utility values and deploys them in the cache space of the edge streaming media server.