Performance issues

[gvarada]

Hi,

I've 10 threads/sinks draining events from a file channel , the source is thrift legacy source. I've a jmeter running 10 threads pumps out about 15000 events in 20 seconds each event is 2k in size. My batch size is 100. It takes over 3 mins to drain these events and it appears that most of time is spent writing to Cassandra. My Cassandra is a 3 node cluster. Any thoughts on where to look or what to do speed up the writes ? Have you done any bench. Marks on the Cassandra sink?

Thanks
Gv