Multimedia Proxy Caching Mechanism for Quality Adaptive Streaming Applications in the Internet

1 Summary Caching is one common technique for enhancing the scalability of media streaming systems. However, current caching techniques are not aware of quality adaptation. To deal with this problem, Rejaie et al. (1999) proposed an adaptive multimedia caching mechanism for layered encoded multimedia streams. This mechanism includes two techniques. The first is a pre-fetch scheme that enables the proxy to improve the quality of cached streams in a demand-driven fashion. The proxy always caches a stream during the first playback. During subsequent playbacks, at any playout time tp, the proxy will pre-fetch any missing segments in the interval of [tp+T, tp+T+δ], called the pre-fetching window. This pre-fetching window will slide as fast as the playout point. Multiple requests of different clients are batched and sent to the server. The server will then send missing segments in decreasing priority. A new pre-fetching request will preempt the previous one. The second technique they proposed is the replacement algorithm. To choose a victim layer, they calculate the whit (weighted hit) for each layer during an interval, called popularity window. Whit is computed by Tplay/TTotal. The popularity of a stream over a popularity window is the sum of its whits during this interval. The last segment of the least popular layer will be replaced first. However, while a stream is being accessed, all its layers are locked and cannot be removed. To evaluate the two above techniques, the authors conducted some simulations in ns. Their simulations showed that the interaction between the replacement and pre-fetching algorithm converges the state of the cache to an efficient state such that the quality of a cached stream is proportional to its popularity and the variations in quality of a cached stream is inversely proportional to its popularity.