Using dynamic segmentation adjustment to improve the performance of streaming proxy servers

As the traffic of Internet dramatically increases in recent years, the proxy server has demonstrated its effectiveness to mitigate the Internet latency. Due to the different characteristics between multimedia and non-multimedia webpage files, more and more proxy servers are dedicated solely to handle multimedia files. Compared with regular webpage files, multimedia files are often transmitted with the technique of streaming so uses can start viewing them as soon as they are reaching the client's site without awaiting the completion of the entire transmission. Those multimedia proxy servers are also referred to as streaming proxy servers. In general, the sizes of multimedia files are much larger than those of regular webpages. It is unlikely that a steaming proxy server can constantly store entire contents of multimedia files in its memory. As a result, the streaming proxy server needs to split individual multimedia files into segments and only stores popular segments in its memory. Researchers had proposed various ways to do the segmentation. However, the sizes of individual segments are often fixed once the segmentation is done. We argue that the sizes of individual segments should vary according to their popularity. A popular segment can have a longer length so the overall performance can be increased accordingly. We propose a novel design, Dynamic Segment Size (DSS), which dynamically adjusts the length of segments by their popularity. We apply our design to a sophisticated algorithm, Adaptive and Lazy Segmentation (ALS), which performs the work of splitting multimedia files into segments and handling memory replacement in a streaming proxy server. Through experimental simulations, our results show that DSS can help ALS increase its byte hit ratio by up to 8.15%.