I have try using push replication 18.104.22.168329 which using event distribution pattern over messaging pattern. When I try updating data in cache around 150 objects per sec for 20 minutes, latency is around 1-2 secs which ok for me. After 20 minutes we stop updating anything in cache but CPU is still high
We have 4 storage nodes in cluster. when it replicate one of them is using CPU around 30-40% all the time, others are around 10%. After 20 minutes, one of them still using CPU around 30-40 % while others are 0%-1%.
Note that I try call suspend() operation in EventChannel through JConsole. CPU is down from 30-40% to 0-1%. I call start() later. CPU do not grow up anymore. So I'm not sure that I have wrong setting or there is a problem in event distribution pattern or messaging pattern?
deep into the event distribution pattern code, OnDistribute() method in AbstractEventChannelController call "ScheduledExecutorService" which schedule task every 1 sec (default) even though there is no event to distribute. This loop seems caused a CPU problem.