SCADIS: Supporting Reliable Scalability in Redis Replication on Demand

摘要

Through replication, Redis has played a primary role in the improvement in the reliability and performance of NoSQL-style database systems in addition to applying added characteristics such as database flexibility and scalability. However, while Redis enables the seamless replication of datasets, Replication becomes difficult when applied to high-volume databases. As a result, effective database scaling becomes a considerable task in large-scale development and maintenance. In this paper, we present a new approach, SCADIS, that alleviates critical problems in reliable Redisreplication-based database scalability. This approach presents three major features: 1) reduced configuration time during the supporting of background and foreground processes; 2) automatic master node failover capabilities; and 3) a high availability service for monitoring Redis clusters. Our experiments upon the proposed cluster deployments demonstrate that SCADIS provides reliable failovers, and effectively scales in and out during write operations over various cluster sizes.