[SPARK-23310] Perf regression introduced by SPARK-21113

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Done
• Affects Version/s: 2.3.0
• Fix Version/s: 2.3.0
• Component/s: Spark Core
• Target Version/s: 2.3.0
• Assignee: Sital Kedia
• Reporter: Yin Huai
• Created: 02/Feb/18 02:10
• Updated: 12/Feb/18 18:35
• Resolved: 05/Feb/18 18:21

Description

While running all TPC-DS queries with SF set to 1000, we noticed that Q95 (https://github.com/databricks/spark-sql-perf/blob/master/src/main/resources/tpcds_2_4/q95.sql) has noticeable regression (11%). After looking into it, we found that the regression was introduced by SPARK-21113. Specially, ReadAheadInputStream gets lock congestion. After setting spark.unsafe.sorter.spill.read.ahead.enabled set to false, the regression disappear and the overall performance of all TPC-DS queries has improved.

 

I am proposing that we set spark.unsafe.sorter.spill.read.ahead.enabled to false by default for Spark 2.3 and re-enable it after addressing the lock congestion issue. 

Issue Links

• is a parent of: SPARK-23366 Improve hot reading path in ReadAheadInputStream
• is caused by: SPARK-21113 Support for read ahead input stream to amortize disk IO cost in the Spill reader
• links to: [Github] Pull Request #20492 (sitalkedia)
• links to: [Github] Pull Request #20514 (ueshin)