I develop WSO2 SI application containing 200+ different queries which process events from single stream. Queries contain multiple conditions, regexes, searches in in-memory tables. A little number of queries writes smth to RDBMS-table. Result of processing of queries is publishing to rabbitmq.
SI is working as standalone daemon, Virtual machine has 20+ CPU cores. htop says that SI process is using 100-300% of CPU (1-3 cores). But it is not enough to process all incoming events (1-2k events per second) - inbound queue (rabbitmq) sometimes grows up to millions of events.
How can i increase utilization of CPU cores? Will partitioning of queries helps me? If no partitions are used, WSO doesn't use parallel processing (for example parallel query )?