A Distributed Storage System For P2P-Vod Service

The online video service, as the fourth largest Internet application besides online music, news and instant messaging, has become the most important medium for people accessing to film, television and other digital content. Peer to Peer Video on Demand (P2P-VoD) service, as the most popular online video service, has large scale of customers. As the development of the Internet infrastructure and video media, more and more customers use VoD service for their entertainment. However, the expending of the customer scales and the increasing of the demand of popular media programs has brought stricter requirement for online video services.The rapid increase of amounts of media data has brought new problems for VoD systems. First of all, the storage of the mass media data causes huge hardware costs for VoD services. Secondly, the availability of the system and the reliability of data should be ensured to keep the normal running. At last, it is better to reduce the system load as much as possible to improve system performance, since large amounts of customers may use the system simultaneously and they need better experience of smooth video playing.The SourceServer of P2P-V6D system is responsible for the storage of the whole data of media programs. The mass concurrent data requests during short periods brings heavy pressure to SourceServer, causing high system load, rapid increase of system resources, decrease of system processing. Moreover, some unexpected system failure may result in failure of SourceServer, system collapse, and even losing data. These failures will affect the normal running of system, causing video playing not smoothly, and the decrease of user experience.With the deepening of information services, more and more data requires for long-term and low-cost storage. Mass data storage and processing has become the most important issue for datacenter. The distributing storage of data and cloud computing technologies quickly become the research focus in IT fields. Hadoop Distributed File System (HDFS) is the basic file system for Apache Hadoop project. It can be built with low-cost common PC cluster, meanwhile provides high reliability and high fault tolerance for the system. It can offer high throughputs of data access for data-intensive applications. HDFS is very suitable for long-term mass storage data sets.This paper proposes a new solution for problems of SourceServer of P2P-VoD system-a reliable and high-performance distributed storage system for P2P-VoD service. This distributed storage system uses HDFS’s scalability and reliability features. In addition, a new function of DataNode local cache for HDFS is designed and developed to improve performance of system. Through building a real experimental environment, we prove the system availability and reliability using a variety of tests methods. Furthermore, we verify that the system performance of SourceServer has been greatly improved.