Performance analysis of request migration in video-on-demand systems

The random early migration (REM) scheme was proposed in our previous work to balance the load of multiple media servers to decrease the average service delay. When an user request arrives, it is randomly directed to a media server that has the designated video content cached on. When the load of this server exceeds a preset threshold, REM is executed by choosing one of its in-service requests and migrating it to another media server with a certain probability, where the exact probability is a function of the service load. We introduce a state matrix representation that stores the service load information of each media server and plays an important role in the determination of migration paths. All possible state matrices can be mapped to a vector space called the state matrix space (SMS). With SMS, we can analyze the performance of VoD systems such as the failure rate and service delay, and these derived results are verified by numerical experiments. It is demonstrated that REM outperforms the normal migration scheme with shorter service delay and lower failure rates.