Decentralized fault tolerant caching with Memcached
Recent changes in web trends not only have increased popularity of web services, but also have vastly increased usage. Popular techniques, such as web-caching, used by content and service providers to provide faster content delivery at user end, is not enough to keep up, due to diversity and dynamic nature of web-contents. Emerging ideas to cache on the server side, for faster content generation and delivery, is currently in use by popular web-service providers. One example of such caching system is Memcached. However, memcached, a distributed high performance caching system, is ineffective in dynamic organization of itself and scaling when required, such as failure or adding more systems on demand. In this theses I have designed and implemented a group membership protocol within a set of Memcached servers, used as a caching pool, that can make the pool aware of its participants. The protocol also helps to dynamically resize itself in case of change of members in the pool. This way in case of both failure and adding more nodes, the pool automatically adjusts itself and publishes the information to the pool clients when adapted to the protocol. I have also shown effectiveness of the protocol by running several test cases.
Das, Bivas, "Decentralized fault tolerant caching with Memcached" (2011). ETD Collection for University of Texas, El Paso. AAI1494339.