Abstract
In this paper we propose a new scheme for scalable video services across the Internet. We envisage the future Internet to be very video intensive where large video and multimedia files (or portions of files) would be accessed over the entire network by a large number of users very frequently. Such large sets of data would contribute to a significant fraction of the entire traffic. Quality of service guarantees would therefore have to be provided for such services. We consider video traffic and show how good service can be provided using our new, scalable and generic notions of video caching and virtual video caching. In order to measure the success of such caching schemes, we introduce the notion of blackouts. A blackout is the break in viewing a video movie between two consecutive clips due to download delay. We define blackout length as the sum of the durations of all blackouts occuring during the service of the video request. We use our generic caching techniques to provide video viewing with low values for average blackout frequencies and blackout lengths; when we use virtual video caching, we get much better performance than what ordinary video caching offers. Our performance studies are based on extensive simulations done by varying parameters such as request rates, cache sizes, queue sizes and packet sizes. Our techniques apply also to large non‐video files that need to be downloaded as streams of smaller chunks in a sequence.