[HIVE-3706] getBoolVar in FileSinkOperator can be optimized

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.10.0
• Fix Version/s: 0.10.0
• Component/s: Query Processor
• Hadoop Flags: Reviewed
• Assignee: Kevin Wilfong
• Reporter: Kevin Wilfong
• Created: 13/Nov/12 18:01
• Updated: 10/Jan/13 19:53
• Resolved: 15/Nov/12 08:41

Description

There's a call to HiveConf.getBoolVar in FileSinkOperator's processOp method. In benchmarks we found this call to be using ~2% of the CPU time on simple queries, e.g. INSERT OVERWRITE TABLE t1 SELECT * FROM t2;

This boolean value, a flag to collect the RawDataSize stat, won't change during the processing of a query, so we can determine it at initialization and store that value, saving that CPU.

Issue Links

• is depended upon by: HIVE-3710 HiveConf.ConfVars.HIVE_STATS_COLLECT_RAWDATASIZE should not be checked in FileSinkOperator