Details
-
Bug
-
Status: Closed
-
Major
-
Resolution: Invalid
-
Product Backlog
-
None
-
None
-
ghx-label-2
Description
Running TPC-DS queries with concurrency of 80 on 1TB scale on a 7 node cluster I'm seeing average throughput of 596 queries per hour on the buffer pool branch versus 681 queries per hour on master. Need to investigate further to figure out why.
It turns out that the buffer-pool build was getting worse throughput because it would run the very long queries through to completion much more often than master.