Yohsuke Ozawa

A Hybrid Event-Processing Architecture based on the Model-driven Approach for High Performance Monitoring

Business activity monitoring (BAM) is a concept for computer systems which support monitoring situations and the performance of business processes as they execute. The monitoring system needs to capture and to process numerous events while correlating them with other events stored in a database. If the database accesses become very frequent, read-only and write-only accesses can achieve high throughputs by applying well known solutions; a result cache and a batch update. However, the event processing of the monitoring includes both simultaneously. It is the case that is hard for the existing solutions to achieve high throughputs. Here, we introduce a new architecture, HM-hybrid, to process events for monitoring efficiently. It is a hybrid of a result cache and a batch update. And our model-driven approach makes it possible for them to work together in situations that are normally regarded as being unsuited for them. It processed 1207 events per second, about 43000 queries per second. It is 6 times better performance over a naive caching architecture in our experiments. The architecture seems useful for monitoring applications where there are many update and select queries.