Improved On-Line Stream Merging: From a Restricted to a General Setting

Stream merging is a promising technique for reducing server bandwidth in video-on-demand systems. There are many heuristics for the problem proposed whose effectiveness has been confirmed empirically. However, it is desirable to prove their effectiveness mathematically. In the pioneering work [2], Bar-Noy and Ladner studied stream merging using competitive analysis. They designed an O(log n)-competitive online scheduler, where n is the totaln umber of stream requests. However, their result is applicable only to systems with large client bandwidth and buffer size. In this paper we design the first on-line scheduler for stream merging in the general setting, in which we lift the large resource requirements, and our scheduler achieves constant competitive ratio.