具有容錯功能的視訊伺服器

Abstract

Fault tolerance requirements for a real-time multimedia server become an important issue in recent years. In a video-on-demand (VOD) environment, disk arrays are often used to support the disk bandwidth requirement. As the number of disks in the video server increases, so does the probability of a disk failure. This can cause serious problems when some disks cannot function normally. Traditionally parity and duplex based approaches provide the capability of recovery but may not guarantee the quality of service. We propose a new scheme of data placement to maintain the quality of service as well as possible in the presence of a single disk failure. In this thesis, we explore the approach of replicating frequently accessed movies to provide high data bandwidth and fault tolerance required in a disk-array-based video server. For those movies that are not so popular, we use parity information to recover the original data. To evaluate the performance of our proposed method, we conduct a series of experiments by simulating the storage and delivery of movies in a VOD system. We compare our method with other approaches in terms of fault tolerance capability after disk failure. Other modules such as admission control and job scheduling algorithm are also studied.