Lazy view maintenance for social networking applications

Abstract

We introduce CAMEL, a lazy view maintenance system for social networking applications on a database server with a distributed memory cache. System administrators can control the throughput of the system by tuning the level of freshness of materialized views. CAMEL employs the existing view maintenance techniques of incremental maintenance, lazy maintenance, and control table. In addition, CAMEL optimizes view maintenance performance by pushing the top-k operation down to before join operations and by constructing a reverse index. We evaluate CAMEL using real data from a mini-blog service. The results show that CAMEL is 6.13 and 11.2 times faster than the method of eager view maintenance while keeping the freshness of materialized views at 66.2% and 38.0%, respectively. © Springer-Verlag Berlin Heidelberg 2010.