Joe McDonnell created IMPALA-15554:
--------------------------------------
Summary: Add ability to run AB performance tests against larger
datasets
Key: IMPALA-15554
URL: https://issues.apache.org/jira/browse/IMPALA-15554
Project: IMPALA
Issue Type: Epic
Components: Infrastructure
Affects Versions: Impala 5.0.0
Reporter: Joe McDonnell
The existing bin/single_node_perf_run.py uses the Impala devenv to compare
performance between two commits on a single machine. This is useful, but it
currently operates at a small scale. Typically, it is testing TPC-H 42 and
TPC-DS 20. Since it is loading data from scratch, larger datasets have higher
overhead in a Jenkins job. This epic tracks work to be able to do higher scale
comparisons using that framework. In particular, it would be useful to have
datasets in S3 with local data cache to avoid dataload time. Isolating the
Impalads using cgroups cpusets would improve consistency.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
