Our application uses HA (prcecise recovery with JMS). it is is a simple EPN: JMS adapter->HA adpater->channel->processor->channel->HA adapter->JMS adapter
All events has to go through the same processor.
We have a throughput problem. It is not meeting the required throughput rate. It seems that the butttleneck is the speed of getting the messages from the WL JMS topic.
If we use multiple JMS adapters to read in parallel (we also use mesage selector in this case to avoid multiple reading of teh same message) then there is no buttleneck in reading the messages from the WL JMS topic. Also there is no buttleneck in the CQL processor, CPU or memory.
The problem that we have using multiple JMS adapters is that the order of the events is not gurenteed so we are looking for a solution that will keep the order and increase throughput.
Any ideas, best practice or reference to documentation will be appriciated.
Edited by: 884793 on Aug 6, 2012 10:25 AM
Edited by: 884793 on Aug 6, 2012 11:02 AM
Edited by: 884793 on Aug 6, 2012 11:03 AM