The solution of Pao Server system in Guangdong Netcom VOD project

1. Project background

With the rapid popularization of the Internet, the number of Internet access has increased by leaps and bounds. At the same time, there is a growing demand for information on the Internet. Gradually not satisfied with the simple text interface and boring MIDI music, and eager to be able to hear the more cordial to see the sound and image. So the network streaming media technology is born with the development of the network, and integrates the television technology, and becomes the rapid rise of a high-tech comprehensive technology. It utilizes the advantages of the network and video technology, completely changed the past to watch the passive way of the program, to achieve the program on demand to watch and any play, set dynamic film and television images, static pictures, sound, text and other information as a whole, to provide users with real-time, interactive, on-demand service on- Streaming media and real-time online events will be a new area and a vibrant market.

As a new way of media and information dissemination, the Internet (Internet) has been gradually developed into every corner of the world, and the Internet is becoming one of the important sources of information, just like the traditional media such as radio, television, newspapers and magazines. The rise and development of the Internet is a great challenge to the cause of radio and television, at the same time, it also brings great development opportunity to the broadcasting and television industry, utilizes the communication technology superiority of the Internet and the flexible communication idea to realize the strong combination and complementation of the radio and TV with the Internet, thus occupying the commanding heights of the modern communication technology and promoting the initiative, is the key to the Internet and the opportunity to win the development of radio and television, but also the only way for the development of radio and TV in the internet age.

With the wide application of broadband technology and the continuous improvement of decompression technology, video playback on the internet has become possible. In recent years, automatically according to the speed of the network transmission of the corresponding image streaming media technology popularization, so that the VOD video playback when the problem can be eradicated, to the Internet to provide a video program broadcast conditions, so that a real sense of the online TV has become possible.

Guangdong Netcom has a wealth of bandwidth resources to Guangzhou IDC Center as the core, to provide 2.5G of export bandwidth, while in Shenzhen, Dongguan, Foshan, there are three centers, network user resources radiation in Guangdong province. To further develop user resources, and at the same time allow existing customers to enjoy a wide range of information services, plans to introduce a large streaming media VOD system for existing 50,000 of online users to provide broadband video information entertainment services, Netcom users as long as through the ordinary PC terminals, through the ordinary ADSL internet, You can see the dynamic movie image clearly and smoothly on demand.

2. Solution System Design-Netcom Practical application Analysis:

In Guangdong netcom application of IP VOD system, to the current network of 50,000 users to consider, according to the general 20% clicks, the current number of concurrent on-demand will be as high as 10,000. In addition, the future with the increase of users, the entire VOD system must also be extended. Therefore, telecom-grade VOD system must have a good expansibility. In order to solve the demand of IP VOD system, it is proposed to adopt modular-distributed system, which is centralized management, storage distribution, load balance distribution, and the preparation of smooth expansion and resource sharing for future system development. In this way, the system can better adapt to the increase of users on demand, reasonable use of network bandwidth, for the system economic benefits, the future expansion of the system, Network Service quality assurance and other aspects bring great benefits.

The basic idea of system design: in Guangzhou center only place WWW server, authentication server, billing server, database server, Management Server, management station and part of streaming media server, centralized management, unified billing, with a unified portal to the whole district users to provide services, users in the central unified certification, and other streaming media server distribution in the vicinity of Shenzhen, Dongguan, Foshan, the formation of VOD server system cascade.

In the central region, streaming media server performance is best, using the high-end four-way Xeon server pt4800r, dual-machine cluster to achieve Media Resource Center services; the surrounding area of the streaming media server recommended cost-effective new Xeon (extended 64-bit) tower PT6510 implementation, and increases as the number of users in the local range increases.

All over the streaming media servers can also be to the entire central region and other surrounding areas of user services, the load balance between the servers, to achieve program dynamic distribution, automatic distribution to solve the high concurrent flow requirements.

3. System working principle:

3.1 Server distribution placement, lifting network backbone possible bottlenecks

If the use of the video server centrally placed in the center, all users on-demand streaming from the center, then, regardless of the center of the server cluster has a good load-balancing mode, can only balance the server itself resource consumption, but from the center to provide services outside the backbone of the network bandwidth is limited, When reaching a certain user on demand (such as 10000), the backbone bandwidth of the central network becomes the bottleneck of the whole system. In the system design, if not to put the server centrally placed in the center for centralized VOD services, but the distribution of video servers in a balanced location, so that the entire region of users on demand will be distributed everywhere, the network load is distributed to different regions of the telecommunications backbone, Then use the load-balancing method of distribution according to the traffic, which will not create bottlenecks in the telecom backbone of any region.

3.2 Storage mode:

In addition to the central region to store all of the chip sources, the system of other video servers in the structure and function of the primary and secondary, are equal, each video server program storage mode is suggested as: the storage space is divided into two parts, part of 90% (this value can be set), as a fixed program storage, and another 10% as a program dynamic distribution. The programs stored in the fixed program storage space (90%) of each server in the system are uploaded by the central administrator, and the programs in each server are different, while the programs in the dynamically distributed space are automatically distributed from other servers as necessary (as determined by the load balancing method). The programs in this part of the space may have the same, and when this part of the space is gone, the system will automatically delete the least-requested programs in the space and continue to buffer the programs that are distributed by other services. There are two major imperatives and advantages:

A, the most resource utilization: Because of the different programs in the fixed program storage space (90%) in all servers, there is no dedicated buffer server or edge server for full distribution or backup, each server has high storage resource utilization (90%).

B, only by dividing the entire storage space into two parts, can we really ensure that the effective storage space always maintain a high utilization rate. We recommend that each server fix the space used for load balancing when the space (such as 100GB) is full, the system automatically statistics which part of the film is on demand at least and delete, leaving room for later buffer program, so that the buffer program to use the storage space within the limits, Rather than unrestricted use, so that the fixed program storage space is not affected and maintain high utilization. If the buffer space is not fixed, then as the buffer program increases, the amount of free space on each server is reduced, and the buffered programs may be buffered to multiple (or even each), and after a while, there may be a buffer in the storage space of all the servers, and perhaps 80% of the programs will be the same, This will cause a great waste of storage resources of the server, and it can bring inconvenience to the management of server resources.

3.3 User Request Service mode

The proposed system is designed by the center (such as the central region of the telecommunications company) to the entire system of unified management, unified construction and provide a unified portal and interface for user Services, all user authentication, billing, program management by the central centralized management. User on-demand first into the center of the WWW server, and then authenticated login, users can be on demand. On demand, the user requests information to play the movie (movie name, the movie's server, movie path, etc. will be sent to the center's WWW Management Server, and the server will analyze the information and decide to send it to the video server where the program is located, which sends the data directly to the user, The client player starts playing, and the client only communicates with the video server during playback and no longer passes through the hub server. Because the programs stored in each server are not the same, so for a program that just uploaded, it only exists on a single server, and before the load-balancing condition (measured in traffic) is reached, all users on-demand for the movie are serviced by this server. When the video on-demand rate reaches certain or the concurrent user exceeds the set value, the load balance and the dynamic distribution of the slice source will be carried out.

3.4 Load Balancing mode and principle

In this system, assume that in the future in Shenzhen, Dongguan, Foshan three cities in accordance with the above way balanced distribution of SA, SB, SC three sets of video servers, three cities in three sets of servers through the telecommunications backbone network interconnection. The system uses the following two ways to load balance:

A, the program is automatically distributed to the nearest user's server: Because in the system, according to the above program distribution, each server program distribution is different, that is, a film in the original only on a server, we assume that the SA server. At the beginning, in Shenzhen, Dongguan, Foshan City, three of all users on demand for the film, are directly from the Shenzhen SA server directly to all users on-demand services, the server will automatically calculate the film by the number of users on demand in different cities. In a certain period of time when a city such as the Dongguan user on demand the number of times the movie reached a certain number (the number can be set), then the server load balancing software will automatically transfer the program copy (distributed) to the SB server in Dongguan, then Dongguan users on the video on demand, are no longer serviced by the Shenzhen SA server, but directly by the SB server for the Dongguan user (the Foshan user's demand is still serviced by the SA server). Similarly, when the Foshan user on demand to reach the set value, the film will automatically be distributed to the Foshan SC server, Foshan users on demand directly from the SC server services. This brings two benefits:

--Load balance: When the program is automatically distributed, all the focus on the film into a number of servers for different user groups to provide services, the centralized load distributed to multiple servers to deal with, to achieve the purpose of load balancing;

--Faster response time: When a server on the higher concurrent flow, the user's on-demand response slower, and after the automatic distribution of the program, the user's request is always the closest to the user's server to respond, so that the fastest for the user service.

B, the program automatically distributed to the smallest load of the server: in the large-scale development of VOD users, concurrent flow is increasing, the video VOD server cluster in the concurrent streaming on-demand peak still can through scheduling software (load balancing calculation) to ensure the efficiency of its video on demand.

When a server, such as the SA server, reaches a critical number of concurrent requests (for example, the concurrent stream is 800), when the concurrency stream reaches 700, the server load Balancing program calculates a portion of the movie with the highest on-demand rate at the same time (such as 5) and determines which server in the entire system has the fewest concurrent users. The 5 movies are then automatically distributed to the server with the smallest load (such as SC), and when new users are on demand for the 5 videos, they are no longer serviced by the SA, but SC provides on-demand services.

In both of these balances, all programs that are distributed from other servers are stored in the fixed buffer above, and when there is no space in the buffer, the server program calculates which of the server buffers, such as 5, has the fewest number of movies on demand, and deletes the movies for future buffering.

4. System Selection and configuration: (Slightly)5. Overall Solution System Advantages:

1, the System support server cascade, to the region as the center, the establishment of film and television channel portal, such as the Guangdong Unicom Unified Portal to the user, the whole region all video server program sharing, resource sharing.

2, the system supports the "centralized on demand, distributed play" function, the user login system can choose the nearest server program to the user to play;

3, the system has dynamic statistics and dynamic distribution of the chip source, according to the user on-demand statistics will be all over the city on demand high rate of the chip source dynamic transmission to the local server;

4, the application of flexible: The entire system can be convenient networking, that is, you can use only one server to provide services, you can also use multiple servers in the system, and to make multiple servers automatically load balance; In this system, the Guangzhou Central area, the Web server, streaming media server, Management Server, etc. together, The use of cluster services to share the enclosure mode, can also be divided into various types of applications to make security better, higher performance.

It has been proven that after planning the products and programmes provided by Pao, Netcom's three resources to achieve a flexible call, not only the entire system load capacity, the realization of the maximum hardware resources, online users increased, the use of better, and a variety of application planning clear, for Netcom to provide online users of the service brought a new upgrade, Greatly promoted the development of Netcom's online business.