COSAR-CQN: An Application Transparent Approach to Cache Consistency

Cache managers speed up the performance of data intensive applications whose workload is dominated by queries. An example is memcached which is in use by very large well-known sites such as Facebook. A key challenge of such systems is how to maintain the state of the cache consistent with that of the database in the presence of updates. With SQL based database management systems (DBMSs) that support query change notification mechanism, one possible approach would require the cache manager to a) register queries used to compute a key-value pair with the DBMS to subscribe for notifications when the query result sets change, and b) delete the cached key-value pair(s) once notified of a change by the DBMS. This approach is independent of how the application updates the database, eliminating application specific software for cache consistency. This reduces the complexity of the application software and minimizes the software development cycle associated with designing, developing, and maintaining software in support of cache consistency. This paper presents the design and implementation of this approach in a cache manager named COSt AwaRe, COSAR. This approach is named Continuous Query change Notification, CQN. It is one of several cache consistency techniques implemented in COSAR. We compare CQN with these alternatives, quantifying its strengths and limitations.