[HIVE-16414] [Hive on Tez] Hive Union queries resource efficiency less on Tez than Mapreduce

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 2.1.0
• Component/s: Tez
• Assignee: Unassigned
• Reporter: Ravi Teja Chilukuri
• Created: 10/Apr/17 11:33
• Updated: 09/May/17 04:44

Description

When a hive union query with the sub queries reading the same table is run in Mapreduce and tez, Mapreduce reads the table only once, no matter how many reads on the same table are present,
but tez reads the same table multiple times in the form of multiple vertices.

If a table is to be read by X mappers,
Tez runs with kX map tasks where k is the number of sub queries reading from the same table and
Mapreduce runs with X mappers no matter how many sub queries are present.

For such union queries, we need to fall back to MR instead of TEZ.

Query:
http://pastebin.com/t6n91u6a

Tez explain plan:
http://pastebin.com/aWwVxhii

MR explain plan:
http://pastebin.com/iDbWwtKR

Issue Links

• is related to: HIVE-16602 Implement shared scans with Tez