Parallel Cache Management Protocol for Static and Dynamic Web Contents

This paper studies Random Request Distribution and Load Balancing Cache protocol (RLC) and Randomly Selected and Limited Look at neighbor (RSLL) cache algorithm based on Byte Access Frequency Factor (BAFF) for internet services with multimedia and dynamic web content. The RSLL cache algorithm considers reducing response time and achieving good load balance among web servers using limitation of look at neighbor web server’s cached contents and random distribution of web client requests. Our cache protocol and algorithm achieve up to 63% gain in performance for response time and 9.3 times more memory hits compared with conventional LRU and LFU algorithms. This approach has good performance gain both in current web client’s request pattern such as non-uniform request pattern and in ideal web client’s request pattern such as uniform request pattern. The RLC cache protocol updates cached dynamic contents completely and consistently in distributed cooperative web servers. Recently, fast response time for web client request has become more and more important in web servers. Therefore, by using this cache protocol and cache algorithm for web servers, we can build scalable web servers supporting very fast response times, high memory hit rate, and excellent load balancing of distributed cooperative web servers.