The research and improvement of distributee caching system Memcached

The Memcached system has no redundancy mechanism. In special cases, there will be a performance degradation of the server and the single-point-of-failure. For these reasons, MS_Memcached, a distributed cache system with master-slave mode, is presented in this paper. It maintains the slaves from the masters with the asynchronous message queue. By using a dynamic feedback load balancing algorithm, the MS_Memcached implements the “vertical” load balancing and solves the single point of failure with an election algorithm. By using Amazon's data partitioning scheme, it facilitates the uniform distribution of data.