Addressing the Read-Performance Impact of Reconfigurations in Replicated Key-Value Stores

Raw data are often orders of magnitude larger than main memory for many applications. As the performance storage devices is still significantly slower memory, systems rely on caching to improve performance. Data replication schemes prevalent in stores high availability and reliability. In such schemes, while updates propagated all replicas (either synchronously or background), reads usually served by only a subset replica group members (e.g., as primary-backup quorum systems). result, non-serving cannot keep their cache state updated; thus, during reconfiguration fail-over action, system suffers from read-performance impact significant amount time due cold-cache misses. our study we observed up 70% hit after cold misses, taking almost 18 minutes some cases fully restore pre-reconfiguration level this article propose mechanism maintain up-to-date read caches across sending hints warm. Thus able seamlessly achieve same even face reorganization. This especially important under read-intensive workloads that common today. Our evaluation shows has benefits reconfigurations, with low periods resource strain. Given its advisory nature, maintenance can be reduced held off if needed periods.