Optimizing a “content-aware” load balancing strategy for shared Web hosting service

FLEX is a new scalable “locality-aware” solution for achieving both load balancing and efficient memory usage on a cluster of machines hosting several Web sites. FLEX allocates the sites to different machines in the cluster based on their traffic characteristics. We propose a set of new methods and algorithms (Simple+, Advanced and Advanced+) to improve the allocation of the Web sites to different machines. New methods show additional performance benefits as compared to the original Simple strategy. Experiments show that FLEX outperforms traditional load-balancing solutions by 50% to 100% (in throughput), even for a four-node cluster. The miss ratio is improved 2-3 times