You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@drill.apache.org by "Aman Sinha (JIRA)" <ji...@apache.org> on 2015/03/07 00:57:39 UTC
[jira] [Created] (DRILL-2402) Current method of combining hash
values can produce skew
Aman Sinha created DRILL-2402:
---------------------------------
Summary: Current method of combining hash values can produce skew
Key: DRILL-2402
URL: https://issues.apache.org/jira/browse/DRILL-2402
Project: Apache Drill
Issue Type: Improvement
Components: Functions - Drill
Affects Versions: 0.8.0
Reporter: Aman Sinha
Assignee: Jacques Nadeau
The current method of combining hash values of multiple columns can produce skew in some cases even though each individual hash function does not produce skew. The combining function is XOR:
{code}
hash(a, b) = XOR (hash(a), hash(b))
{code}
The above result will be 0 for all rows where a = b, so hash(a) = hash(b). This will clearly create severe skew and affects the performance of queries that do HashAggregate based group-by on {a, b} or a HashJoin .on both columns.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)