In this thesis, we propose two methods to improve the performance of a shared file system based web server cluster by increasing the memory cache hit ratio.
One is Cache-Aware Request Distribution(CARD) strategy which distributes incoming web requests in a way to increase the cache hit ratio of back-end server clusters. CARD uses the information of a lock server in a shared file system in order to increase cache hit ratio. This information denotes the recent access user of correspoding data. We implement CARD in a dispatcher of a shared file system based web server cluster. The other is a light-weight cooperative caching algorithm which does not require messages between clients and global cache manager when a block is replaced. This algorithm is compared with previous cooperative caching algorithms through simu-lations. The experimental results show that the performance of the web server cluster with cooperative caching scheme is scalable and CARD increases the cache hit ratio of back-end servers, thereby increasing the overall performance.
