Alex Behm has posted comments on this change.

Change subject: IMPALA-5612: join inversion should factor in parallelism
......................................................................


Patch Set 5:

(7 comments)

Nice!

http://gerrit.cloudera.org:8080/#/c/7351/5/fe/src/main/java/org/apache/impala/planner/Planner.java
File fe/src/main/java/org/apache/impala/planner/Planner.java:

Line 451:    * inversion.
Returns false if any join input is missing relevant stats.


Line 463:    * - the join strategy is PARTITIONED and rows are distributed 
evenly.
I suggest you add a quick note why BROADCAST is not particularly relevant. 
Otherwise, this may seem like a big unexplained assumption.


Line 473:    * The estimated cost of a hash join before and after inversion, 
measured in an
estimated per-host cost


Line 482:    * We choose b = 10 and C = 5 empirically because it seems to give 
reasonable
You had this great spreadsheet over a wide range of inputs. Do you mind dumping 
that table to the upstream JIRA so we can see how this model behaves? 
(Alternatively, just add the link to the Google Spreadsheet.)


Line 491:   private boolean isInvertedJoinCheaper(boolean isLocalPlan, JoinNode 
joinNode) {
nit: flip args to make them consistent with invertJoins()


Line 509:     final long CONSTANT_COST_PER_ROW = 5;
CONSTANT_COST_PER_BYTE?


Line 518:       LOG.trace("lhsCard " + lhsCard + " lhsBytes " + lhsBytes +
Also log the tblRefIds_ and/or tupleIds_ or the join, otherwise it will be 
impossible to know which join this is.


-- 
To view, visit http://gerrit.cloudera.org:8080/7351
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: Icacea4565ce25ef15aaab014684c9440dd501d4e
Gerrit-PatchSet: 5
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-Owner: Tim Armstrong <tarmstr...@cloudera.com>
Gerrit-Reviewer: Alex Behm <alex.b...@cloudera.com>
Gerrit-Reviewer: Bharath Vissapragada <bhara...@cloudera.com>
Gerrit-Reviewer: Mostafa Mokhtar <mmokh...@cloudera.com>
Gerrit-Reviewer: Tim Armstrong <tarmstr...@cloudera.com>
Gerrit-HasComments: Yes

Reply via email to