A Comparative Evaluation of Cache Strategies for Elastic Caching Platforms

Abstract

With the rapid development of cloud computing, traditional TP applications are evloving into the Extreme Transaction Processing (XTP) applications which are characterized by exceptionally demanding performance, scalability, availability, security, manageability and dependability require requirements, elastic caching platforms (ECPs) are introduced to help meet these requirements. Three popular cache strategies for ECPs have been proposed, say replicated strategy, partitioned strategy and near strategy. According to our investigations, many ECPs support multiple cache strategies. In this paper, we evaluate the impact of the three cache strategies using the TPC-W benchmark. To the best of our knowledge, this paper is the first evaluation of distributed cache strategies for ECPs. The main contribution of this work is guidelines that could help system administrators decide effectively which cache strategy would perform better under different conditions. Our work shows that the selection of the best cache strategy is related with workload patterns, cluster size and the number of concurrent users. We also find that four important metrics (number of "get" operations, message throughput, get/put ratio, and cache hit rate) could be used to help characterize the current condition.