Distributed memory stores (e.g. Memcached and Redis) are widely used to speed up user request processing. The unbalanced workload on memory stores is commonly seen in industry applications (e.g. Facebook), which can seriously influence the scalability of supporting a large number of requests. To address this issue, replication is considered as an effective approach. However, existing work does not well solve the problems of replica management overhead and flexible configuration of data consistency. Therefore, The main goal of this topic is to propose a decentralized workload balance framework to improve the scalability of memory stores by solving the hotspot problem.