Active Caching for KVS Dynamic Scaling

Load size to a service in the Internet remarkably changes every hour. Thus, it is expected that service system scale should be changed dynamically according to load size. KVS (Key-Value Store) is scalable DBMS (database management system) and widely used in large-scale Internet services. In this paper, we focus on Cassandra, a popular open-source KVS implementation, and discuss methods for improving dynamic scaling performance. First, we evaluate node joining time, which is time to complete adding a node to a running KVS system, and show that its bottleneck process is disk I/O in the existing nodes. Second, we analyze disk accesses in the bottleneck nodes and point it out that some heavily accessed files cause a large number of disk accesses. Third, we propose two methods for decreasing node joining time. One method is to reduce disk accesses significantly by keeping the heavily accessed file in the page cache. The other method is to increase sequential I/O performance by enlarging time slice of I/O scheduler. Lastly, we evaluate our methods and demonstrate that our method can improve scaling-out performance of Cassandra.