Providing Distributed On - Demand Video

| In an on-demand video system, a number of repository servers store all the video contents of interest to a large pool of geographically distributed users. In order to achieve higher streaming capacity and lower long-haul transmission cost, a distributed servers architecture can be used, in which multiple local servers are placed close to the users and cache the video contents dynamically according to their local demands. We consider in this paper that the repository delivers video contents to the local servers using multicasting. Upon the arrival of a request the local server joins the ongoing multicast streams of the movie and caches the data (i.e., \latches" onto the streams), while a transient network stream is allocated to supply the missing (beginning) portion of the movie due to the time-lag between the request and the multicast streams. After a short while, all the streams will \merge" into one stream. Such scheme is hence called \latching and merging." We study the required number of streams and local buuer for this scheme, the effect of the maximum number of streams that a local server can latch onto, and its advantages in providing distributed on-demand video services.