impala-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Thomas Tauber-Marshall (Code Review)" <ger...@cloudera.org>
Subject [Impala-ASF-CR] IMPALA-4252: Min-max runtime filters for Kudu
Date Fri, 20 Oct 2017 21:20:05 GMT
Thomas Tauber-Marshall has posted comments on this change. ( http://gerrit.cloudera.org:8080/7793
)

Change subject: IMPALA-4252: Min-max runtime filters for Kudu
......................................................................


Patch Set 7:

Perf results:

 +------------+----------+--------------------+--------+-------------+------------+-----------+----------------+-------------+-------+
 | Workload   | Query    | File Format        | Avg(s) | Base Avg(s) | Delta(Avg) | StdDev(%)
| Base StdDev(%) | Num Clients | Iters |
 +------------+----------+--------------------+--------+-------------+------------+-----------+----------------+-------------+-------+
 | TPCH(_100) | TPCH-Q20 | kudu / none / none | 7.04   | 6.10        |   +15.45%  |   3.51%
  |   3.28%        | 1           | 4     |
 | TPCH(_100) | TPCH-Q8  | kudu / none / none | 8.09   | 7.38        |   +9.67%   |   1.11%
  |   2.57%        | 1           | 4     |
 | TPCH(_100) | TPCH-Q17 | kudu / none / none | 19.13  | 17.50       |   +9.33%   |   3.91%
  |   4.34%        | 1           | 4     |
 | TPCH(_100) | TPCH-Q22 | kudu / none / none | 2.71   | 2.56        |   +6.00%   |   0.90%
  |   4.11%        | 1           | 4     |  
 | TPCH(_100) | TPCH-Q16 | kudu / none / none | 4.23   | 4.01        |   +5.45%   |   4.13%
  |   0.53%        | 1           | 4     |
 | TPCH(_100) | TPCH-Q1  | kudu / none / none | 10.87  | 10.55       |   +3.05%   |   1.06%
  |   4.61%        | 1           | 4     |  
 | TPCH(_100) | TPCH-Q18 | kudu / none / none | 24.93  | 24.64       |   +1.18%   |   3.44%
  |   1.29%        | 1           | 4     |
 | TPCH(_100) | TPCH-Q6  | kudu / none / none | 2.27   | 2.26        |   +0.77%   |   0.87%
  |   1.35%        | 1           | 4     |  
 | TPCH(_100) | TPCH-Q7  | kudu / none / none | 18.55  | 18.43       |   +0.66%   |   3.49%
  |   1.40%        | 1           | 4     |
 | TPCH(_100) | TPCH-Q3  | kudu / none / none | 10.48  | 10.43       |   +0.47%   |   0.98%
  |   1.43%        | 1           | 4     |  
 | TPCH(_100) | TPCH-Q4  | kudu / none / none | 15.94  | 15.90       |   +0.27%   |   2.07%
  |   1.26%        | 1           | 4     |
 | TPCH(_100) | TPCH-Q15 | kudu / none / none | 6.01   | 6.00        |   +0.13%   |   2.57%
  |   1.76%        | 1           | 4     |  
 | TPCH(_100) | TPCH-Q19 | kudu / none / none | 8.02   | 8.07        |   -0.63%   |   3.79%
  |   2.33%        | 1           | 4     |
 | TPCH(_100) | TPCH-Q2  | kudu / none / none | 4.07   | 4.10        |   -0.66%   |   5.74%
  |   1.90%        | 1           | 4     |  
 | TPCH(_100) | TPCH-Q9  | kudu / none / none | 18.00  | 18.36       |   -1.96%   |   5.78%
  |   1.17%        | 1           | 4     |
 | TPCH(_100) | TPCH-Q13 | kudu / none / none | 11.44  | 11.67       |   -1.99%   |   3.58%
  |   1.12%        | 1           | 4     |  
 | TPCH(_100) | TPCH-Q12 | kudu / none / none | 6.40   | 6.56        |   -2.37%   |   2.20%
  |   2.28%        | 1           | 4     |
 | TPCH(_100) | TPCH-Q10 | kudu / none / none | 5.42   | 5.60        |   -3.17%   |   1.90%
  |   2.60%        | 1           | 4     |  
 | TPCH(_100) | TPCH-Q5  | kudu / none / none | 10.53  | 11.07       |   -4.87%   |   5.59%
  |   1.44%        | 1           | 4     |
 | TPCH(_100) | TPCH-Q21 | kudu / none / none | 51.31  | 54.29       |   -5.48%   |   2.18%
  |   2.39%        | 1           | 4     |  
 | TPCH(_100) | TPCH-Q14 | kudu / none / none | 4.76   | 7.27        | I -34.49%  |   0.87%
  |   0.54%        | 1           | 4     |
 | TPCH(_100) | TPCH-Q11 | kudu / none / none | 1.27   | 2.56        | I -50.18%  |   2.62%
  |   4.17%        | 1           | 4     |
 +------------+----------+--------------------+--------+-------------+------------+-----------+----------------+-------------+-------+

For the queries that regressed somewhat - Q20,8,17 - the filters simply aren't very selective.
We might consider a path to disable a filter that we see in not selective the way we do with
bloom filters once Kudu exposes stats are filter effectiveness.


-- 
To view, visit http://gerrit.cloudera.org:8080/7793
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: I02bad890f5b5f78388a3041bf38f89369b5e2f1c
Gerrit-Change-Number: 7793
Gerrit-PatchSet: 7
Gerrit-Owner: Thomas Tauber-Marshall <tmarshall@cloudera.com>
Gerrit-Reviewer: Anonymous Coward #345
Gerrit-Reviewer: Lars Volker <lv@cloudera.com>
Gerrit-Reviewer: Matthew Jacobs <mjacobs@apache.org>
Gerrit-Reviewer: Michael Ho <kwho@cloudera.com>
Gerrit-Reviewer: Mostafa Mokhtar <mmokhtar@cloudera.com>
Gerrit-Reviewer: Thomas Tauber-Marshall <tmarshall@cloudera.com>
Gerrit-Reviewer: Tim Armstrong <tarmstrong@cloudera.com>
Gerrit-Comment-Date: Fri, 20 Oct 2017 21:20:05 +0000
Gerrit-HasComments: No

Mime
  • Unnamed multipart/alternative (inline, 8-Bit, 0 bytes)
View raw message