Details
-
Task
-
Resolution: Won't Fix
-
Major
-
4.5.0
-
None
-
None
Description
This ticket was created based on Cihan's request.
This is something that puzzled us:
For MOI, stale=false, we have 720 n1qlworkers. 80% Query latency is 40ms, see chart here:
http://cbmonitor.sc.couchbase.com/reports/html/?snapshot=hera_450-2488-enterprise_490_access. Query throughput is 13000.
For RYOW, we have 1200 n1qlworkers. Each n1qlworkers wait 10ms before issuing queries. Query latency is much lower than stale=false. But we only get query throughput of 5481.
Based on the theory, assume that half of the workers are sleeping during any time, we should have 600 n1qlworkers doing query at any time and this should bring the throughput close to 13000.
This task opens to track this effort and to see if we need to change the perf test.
Attachments
Issue Links
- relates to
-
MB-19489 RC=0x3B[HTTP Operation failed. for self.cb.client.n1ql_query(nq) in Q3 RYOW MOI/Q2 GSI tests
- Closed