Impacts of delayed replication on the key-value store

Recently, a key-value store has been used to improve the throughput and latency performance in distributed computing environments. However, some or all of the data must be processed in the memory to guarantee the performance if the data workload increases greatly. However, there is a bottleneck because of the data replication overheads incurred by the memory to guarantee availability. In this study, we implemented and evaluated the delayed replication of data in the memory to reduce the overheads of the key-value store, which stores and manages data in the memory. The results of the experiments showed that the performance was improved by 11% after applying delayed replication.