A Decentralized Load Balancing Architecture for Cache System

Cache technology has emerged as a promising approach for distributed web systems to mitigate the web server's data access performance bottleneck. In this paper, we propose a decentralized load balancing architecture. The cache system takes advantage of the decentralized architecture for providing scalability and high availability. It can provide content requested by a user if it is cached in any of the cache servers in the system. Moreover, it can reduce the network traffic for web servers if the requested contents are cached in the system. We also propose a dynamic load balancing algorithm called Adaptive Dynamic Algorithm to dispatch requested contents that are not cached in the cache system to different cache servers. In Adaptive Dynamic Algorithm, we combine load metrics (CPU, memory) and number of alive requests to predict the real load of cache server. The balance strategy prevents the cache servers to be skew, improves cache hit rate, and reduces response time effectively. Experimental results support that our proposed load balancing algorithm is faster than Round Robin and Least Connection, and the content hit rate is higher than Round Robin and Least Connection too.