CN104683318A - Edge streaming media server caching selection method and edge streaming media server caching selection system - Google Patents

Abstract

The invention relates to an edge streaming media server caching selection method, which includes the following steps: a plurality of users are grouped into a plurality of user classes according to the respective preferences of the users; the statistics of the intensity of each user class and the preferences of each user class for movies are collected; the intensity of each user class is the sum of the intensities of all the users in the user class, while the intensities of the users are the different influences of the users on the caching decision-making of a provider; according to the preference degree of each user class for the movies and the intensity of each user class, the utilities of the movies are calculated; the movies with the relatively greater utility values are chosen to be deployed into the cache space of an edge streaming media server.

Description

A kind of edge streaming server buffer memory system of selection and system

Technical field

The present invention relates to network communication field, particularly the system of selection of a kind of edge streaming server buffer memory and system.

Background technology

In order to solve the performance bottleneck problem of C/S architecture flow media system, and cut operating costs, content distributing network (Content Deliver Network, CDN) is able to extensive use.CDN places a large amount of edge cache server at network edge.The content of their buffer memorys can directly serve user's order request, avoids user with the data throughput between backbone network, reaches the object reducing data transfer delay, smooth transfer fluctuation, reduction backbone network flow.

For business such as digital interactive televisions, each coverage of network edge also deploys streaming server.Described streaming server is in the centre position of user and CDN, and its effect comprises: the order request of proxy user, and obtain video data from CDN, the form that fluidization treatment becomes IP-QAM to support, is pushed to user.Streaming server is usually very near with the distance of user, for making full use of this advantage, streaming server usually also can possess proxy caching function, store the CDN video file once obtained, directly serve follow-up identical order request, further improvement, to the service quality of user, alleviates the burden of CDN.

Because video file volume is usually comparatively large, no matter is CDN edge cache server, or supports the streaming server of proxy caching function, all face the pressure in memory capacity.Therefore, a vital problem how reasonably to select the buffer memory of caching server to dispose content, effectively utilizes its limited memory space, to give full play to the effect of buffer memory, improve integrity service performance.

Summary of the invention

To the object of the invention is to overcome in prior art and the buffer memory of choose reasonable caching server cannot dispose the defect of content, thus the system of selection of a kind of edge streaming server buffer memory and system are provided.

To achieve these goals, the invention provides the system of selection of a kind of edge streaming server buffer memory, comprising:

Step 1), multiple user is aggregated into some user class by user's hobby separately;

Step 2), statistic procedure 1) intensity of each user class that obtains, and each user class is to the preference of film; Wherein, the intensity of user class is each user's intensity sum in user class, and the Different Effects power that described user's intensity has for the cache decision of user to provider;

Step 3), according to film by the preference of each user class, and the intensity of each user class, calculates the effectiveness of film;

Step 4), choose the larger film of value of utility and dispose in the spatial cache of edge streaming server.

In technique scheme, described step 1) comprises:

Step 1-1), according to user to the viewing time of a certain film and viewing number of times definition user to the preference of this film;

Step 1-2), be that film adds label according to film attribute, according to described label, film is divided into film class;

Step 1-3), to be obtained the preference value of user to film class belonging to this film by the preference of user to a certain film;

Step 1-4), according to user, to user, cluster is carried out to the preference of each film class, obtain some user class.

In technique scheme, described step 2) comprising:

Step 2-1), to perform the active degree of the frequent degree of program request operation and the viewing Time Calculation user of user according to user;

Step 2-2), setting user service class;

Step 2-3), calculate user's intensity according to the active degree of user and service class;

Step 2-4), by user's Strength co-mputation user class intensity;

Step 2-5), in a certain user class, quantize the recent active degree of user;

Step 2-6), with step 2-5) user's active degree of obtaining is weight, weighs the preference of this user class to a certain film.

In technique scheme, in described step 3), the effectiveness of described film is calculated by following manner: with user class intensity for weight, does weighted sum to the preference value of each user class to described film.

Present invention also offers a kind of edge streaming server buffer memory selective system, comprising: user class aggregation module, user class intensity and user class preference generation module, film utility computing module and deployment module; Wherein,

Multiple user is aggregated into some user class by user's hobby separately by described user class aggregation module;

The intensity of each user class that described user class intensity and user class preference generation module counting user Type of Collective module obtain, and each user class is to the preference of film; Wherein, the intensity of user class is each user's intensity sum in user class, and the Different Effects power that described user's intensity has for the cache decision of user to provider;

Described film utility computing module is according to the preference of film by each user class, and the intensity of each user class, calculates the effectiveness of film;

Described deployment module is chosen the larger film of value of utility and is disposed in the spatial cache of edge streaming server.

The invention has the advantages that:

The present invention utilizes cluster and proposed algorithm to draw one's respective area user preference more accurately, disposes foundation, add the accuracy of judgement as buffer memory.

Accompanying drawing explanation

Fig. 1 is the flow chart of edge streaming server buffer memory of the present invention system of selection.

Embodiment

Now the invention will be further described by reference to the accompanying drawings.

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

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

Step 1), multiple user is aggregated into some user class by user's hobby separately.

In order to realize the service of distinction in the application, need the preference of excavating each user from the program request history of user.But the Limited information that the program request behavior of unique user individuality comprises, therefore different users should be gathered in some user class the preference of each film according to it, then utilize collaborative filtering, each user class be analyzed to the importance of each film.

This step specifically comprises:

Step 1-1), first, user u is defined as the preference of film v:

Wherein, n _u, T _urepresent that user u in a period of time is to the video-on-demand times of v, viewing time respectively, and total video-on-demand times of user u, the viewing time.

User u is to the preference of film v constitute the bivector of user, preference value disappearance item (i.e. the not viewed a certain film of user u) wherein remembers zero, or is predicted by SVD model.

Step 1-2), generate film class.Select some film attributes to add label to movie object, optional tag attributes comprises subject matter, show time, related person etc.Be included into by the film of shared label value in a film class, same film can enter multiple film class, and accurate and various tag along sort is the basis of user clustering.

Step 1-3), to be obtained the preference value of user to film class belonging to this film by the preference of user to a certain film.

The preference value of user to a certain film class is the preference sum to such all film.

Suppose that the preference value of user u to a certain label t is: then the preference vector of user may be defined as $P_u=({\mathrm{\ρ}}_u^1,{\mathrm{\ρ}}_u^2,\·\·\·,{\mathrm{\ρ}}_u^t,\·\·\·).$

Step 1-4), last, according to the preference of user to each film class, use K-Means algorithm to user clustering, obtain several user class.

Step 2), add up the intensity of each user class, and user class is to the preference of film.

The Different Effects power that the cache decision of user to provider has is called user's intensity.The definition of user's intensity is according to being that the user class of service is higher, and more active, its intensity is larger.Each user class has different quantity and importance because of comprised user, also presents different importance, is called user class intensity, is user's intensity sum in user class.Wherein, the viewing time two of frequent degree and user that the active degree of user performs program request operation by user, because usually weighing, is the weighted sum of the two; The user class of service depends on the non-technical strategy of service provider, is the important evidence of distribution services resource, and more important user has higher user class numerical value.

User class, to the preference of a certain film, to be defined as in class user to the weighted sum of the preference of this film, with user's active degree for weight.

This step specifically comprises:

Step 2-1), calculate the active degree of user.

The viewing time two of frequent degree and user that the active degree of user can perform program request operation by user is because usually weighing.

Wherein, N _u, T _u, N _u, T _urepresent video-on-demand times, the viewing time of user u in a period of time respectively, and user collects total video-on-demand times, the viewing time of U, the α factor belongs to real number interval [0,1], for regulating the relative weighting of two factors.

Step 2-2), setting user service class.The user class of service depends on the non-technical strategy of service provider, is the important evidence of distribution services resource.Availablely simply user is divided into member user UM and domestic consumer UN.

Step 2-3), calculate user intensity.

The strength definition of user u is:

$\widetilde{i_u}=\{\begin{array}{cc}(\mathrm{\β}+a_u^U)& u\∈U_N\\ \mathrm{\γ}(\mathrm{\β}+a_u^U)& u\∈U_M\end{array}\mathrm{\γ}>1,\mathrm{\β}>0$

Wherein, γ is for distinguishing domestic consumer and member user, and β itself does not have particular meaning, adds that β is just zero in order to avoid this.

User's intensity can do normalized:

$i_u=\frac{\widetilde{i_u}}{{\mathrm{\Σ}}^{\tilde{u}\∈U}\widetilde{i_{\stackrel{~}{u}}}}$

Through Simulation Evaluation, step 2-1) in β, γ in involved α and this step get 0.4 respectively, 1, when 128, method of the present invention has very high cache hit rate.

Step 2-4), calculate user class intensity.

User class intensity is each user's intensity sum in user class, and namely the strength definition of user class c is

Step 2-5), in each user class c, the active degree recent to user quantizes:

$a_u^c=\frac{N_u}{{\mathrm{\Σ}}^{\stackrel{\·}{u}\∈c}{N}_{\stackrel{\·}{u}}}.$

Wherein, ü represents the arbitrary user in user class c.

Step 2-6), with step 2-5) user's active degree of obtaining is weight, weighs user class c to the preference of film v:

$p_c^v=\stackrel{u\∈c}{\mathrm{\Σ}}{p}_u^v{a}_u^c.$

Step 3), according to step 2) each film of calculating by the preference of each user class and the intensity of each user class, calculate the effectiveness of film.

The effectiveness of film is defined as the weighted sum of each user class to the preference value of this film, with user class intensity for weight.Its computing formula is as follows:

${\mathrm{\ψ}}_v=\stackrel{c}{\mathrm{\Σ}}{s}_c{p}_c^v$

Step 4), choose the larger film of value of utility, dispose in limited spatial cache.

Present invention also offers the edge streaming server buffer memory selective system corresponding with aforementioned system of selection, this system comprises: user class aggregation module, user class intensity and user class preference generation module, film utility computing module and deployment module; Wherein,

Multiple user is aggregated into some user class by user's hobby separately by described user class aggregation module;

The intensity of each user class that described user class intensity and user class preference generation module counting user Type of Collective module obtain, and each user class is to the preference of film; Wherein, the intensity of user class is each user's intensity sum in user class, and the Different Effects power that described user's intensity has for the cache decision of user to provider;

Described film utility computing module is according to the preference of film by each user class, and the intensity of each user class, calculates the effectiveness of film;

Described deployment module is chosen the larger film of value of utility and is disposed in the spatial cache of edge streaming server.

Method and system of the present invention utilizes cluster and proposed algorithm to draw one's respective area user preference more accurately, disposes foundation, add the accuracy of judgement as buffer memory.

It should be noted last that, above embodiment is only in order to illustrate technical scheme of the present invention and unrestricted.Although with reference to embodiment to invention has been detailed description, those of ordinary skill in the art is to be understood that, modify to technical scheme of the present invention or equivalent replacement, do not depart from the spirit and scope of technical solution of the present invention, it all should be encompassed in the middle of right of the present invention.

Claims (5)

1. an edge streaming server buffer memory system of selection, comprising:

Step 1), multiple user is aggregated into some user class by user's hobby separately;

Step 2), statistic procedure 1) intensity of each user class that obtains, and each user class is to the preference of film; Wherein, the intensity of user class is each user's intensity sum in user class, and the Different Effects power that described user's intensity has for the cache decision of user to provider;

Step 3), according to film by the preference of each user class, and the intensity of each user class, calculates the effectiveness of film;

Step 4), choose the larger film of value of utility and dispose in the spatial cache of edge streaming server.

2. edge streaming server buffer memory according to claim 1 system of selection, is characterized in that, described step 1) comprises:

Step 1-1), according to user to the viewing time of a certain film and viewing number of times definition user to the preference of this film;

Step 1-2), be that film adds label according to film attribute, according to described label, film is divided into film class;

Step 1-3), to be obtained the preference value of user to film class belonging to this film by the preference of user to a certain film;

Step 1-4), according to user, to user, cluster is carried out to the preference of each film class, obtain some user class.

3. edge streaming server buffer memory according to claim 1 system of selection, is characterized in that, described step 2) comprising:

Step 2-1), to perform the active degree of the frequent degree of program request operation and the viewing Time Calculation user of user according to user;

Step 2-2), setting user service class;

Step 2-3), calculate user's intensity according to the active degree of user and service class;

Step 2-4), by user's Strength co-mputation user class intensity;

Step 2-5), in a certain user class, quantize the recent active degree of user;

Step 2-6), with step 2-5) user's active degree of obtaining is weight, weighs the preference of this user class to a certain film.

4. edge streaming server buffer memory according to claim 1 system of selection, it is characterized in that, in described step 3), the effectiveness of described film is calculated by following manner: with user class intensity for weight, does weighted sum to the preference value of each user class to described film.

5. an edge streaming server buffer memory selective system, is characterized in that, comprising: user class aggregation module, user class intensity and user class preference generation module, film utility computing module and deployment module; Wherein,

Multiple user is aggregated into some user class by user's hobby separately by described user class aggregation module;

The intensity of each user class that described user class intensity and user class preference generation module counting user Type of Collective module obtain, and each user class is to the preference of film; Wherein, the intensity of user class is each user's intensity sum in user class, and the Different Effects power that described user's intensity has for the cache decision of user to provider;

Described film utility computing module is according to the preference of film by each user class, and the intensity of each user class, calculates the effectiveness of film;

Described deployment module is chosen the larger film of value of utility and is disposed in the spatial cache of edge streaming server.