lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Yonik Seeley (JIRA)" <j...@apache.org>
Subject [jira] Commented: (LUCENE-1997) Explore performance of multi-PQ vs single-PQ sorting API
Date Thu, 29 Oct 2009 16:22:59 GMT

    [ https://issues.apache.org/jira/browse/LUCENE-1997?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12771466#action_12771466
] 

Yonik Seeley commented on LUCENE-1997:
--------------------------------------

Here's some more mud to help clear the water ;-)  This is with the latest JDK7 - tested twice
to be sure, and all results were within .5 percentile points of eachother.

Linux odin 2.6.28-16-generic #55-Ubuntu SMP Tue Oct 20 19:48:32 UTC 2009 x86_64 GNU/Linux
Java(TM) SE Runtime Environment (build 1.7.0-ea-b74) (Oct 15 2009)
Phenom II x4 3GHz (dynamic freq scaling turned off) 

||Source||Seg size||Query||Tot hits||Sort||Top N||QPS old||QPS new||Pct change||
|random|balanced|<all>|5000000|rand int|10|28.02|18.86|{color:red}-32.7%{color}|
|random|balanced|<all>|5000000|rand int|25|27.93|18.80|{color:red}-32.7%{color}|
|random|balanced|<all>|5000000|rand int|50|23.89|21.77|{color:red}-8.9%{color}|
|random|balanced|<all>|5000000|rand int|100|23.74|21.21|{color:red}-10.7%{color}|
|random|balanced|<all>|5000000|rand int|500|22.92|17.30|{color:red}-24.5%{color}|
|random|balanced|<all>|5000000|rand int|1000|21.99|14.64|{color:red}-33.4%{color}|
|random|balanced|<all>|5000000|rand string|10|23.63|20.58|{color:red}-12.9%{color}|
|random|balanced|<all>|5000000|rand string|25|22.74|20.42|{color:red}-10.2%{color}|
|random|balanced|<all>|5000000|rand string|50|16.88|21.93|{color:green}29.9%{color}|
|random|balanced|<all>|5000000|rand string|100|19.32|21.42|{color:green}10.9%{color}|
|random|balanced|<all>|5000000|rand string|500|18.58|18.14|{color:red}-2.4%{color}|
|random|balanced|<all>|5000000|rand string|1000|18.08|15.25|{color:red}-15.7%{color}|
|random|balanced|<all>|5000000|country|10|23.89|20.70|{color:red}-13.4%{color}|
|random|balanced|<all>|5000000|country|25|22.59|20.58|{color:red}-8.9%{color}|
|random|balanced|<all>|5000000|country|50|16.84|22.04|{color:green}30.9%{color}|
|random|balanced|<all>|5000000|country|100|16.68|21.71|{color:green}30.2%{color}|
|random|balanced|<all>|5000000|country|500|19.65|18.60|{color:red}-5.3%{color}|
|random|balanced|<all>|5000000|country|1000|17.70|15.48|{color:red}-12.5%{color}|
|random|log|<all>|5000000|rand int|10|28.31|18.94|{color:red}-33.1%{color}|
|random|log|<all>|5000000|rand int|25|23.75|22.09|{color:red}-7.0%{color}|
|random|log|<all>|5000000|rand int|50|23.99|21.90|{color:red}-8.7%{color}|
|random|log|<all>|5000000|rand int|100|23.75|21.47|{color:red}-9.6%{color}|
|random|log|<all>|5000000|rand int|500|22.83|18.41|{color:red}-19.4%{color}|
|random|log|<all>|5000000|rand int|1000|21.99|15.96|{color:red}-27.4%{color}|
|random|log|<all>|5000000|rand string|10|22.92|20.61|{color:red}-10.1%{color}|
|random|log|<all>|5000000|rand string|25|23.36|22.27|{color:red}-4.7%{color}|
|random|log|<all>|5000000|rand string|50|16.96|22.12|{color:green}30.4%{color}|
|random|log|<all>|5000000|rand string|100|19.61|21.59|{color:green}10.1%{color}|
|random|log|<all>|5000000|rand string|500|18.02|19.03|{color:green}5.6%{color}|
|random|log|<all>|5000000|rand string|1000|18.54|16.51|{color:red}-10.9%{color}|
|random|log|<all>|5000000|country|10|24.32|20.65|{color:red}-15.1%{color}|
|random|log|<all>|5000000|country|25|23.46|20.72|{color:red}-11.7%{color}|
|random|log|<all>|5000000|country|50|22.71|20.62|{color:red}-9.2%{color}|
|random|log|<all>|5000000|country|100|16.78|21.78|{color:green}29.8%{color}|
|random|log|<all>|5000000|country|500|19.14|19.22|{color:green}0.4%{color}|
|random|log|<all>|5000000|country|1000|17.61|16.79|{color:red}-4.7%{color}|


> Explore performance of multi-PQ vs single-PQ sorting API
> --------------------------------------------------------
>
>                 Key: LUCENE-1997
>                 URL: https://issues.apache.org/jira/browse/LUCENE-1997
>             Project: Lucene - Java
>          Issue Type: Improvement
>          Components: Search
>    Affects Versions: 2.9
>            Reporter: Michael McCandless
>            Assignee: Michael McCandless
>         Attachments: LUCENE-1997.patch, LUCENE-1997.patch, LUCENE-1997.patch, LUCENE-1997.patch,
LUCENE-1997.patch, LUCENE-1997.patch, LUCENE-1997.patch, LUCENE-1997.patch
>
>
> Spinoff from recent "lucene 2.9 sorting algorithm" thread on java-dev,
> where a simpler (non-segment-based) comparator API is proposed that
> gathers results into multiple PQs (one per segment) and then merges
> them in the end.
> I started from John's multi-PQ code and worked it into
> contrib/benchmark so that we could run perf tests.  Then I generified
> the Python script I use for running search benchmarks (in
> contrib/benchmark/sortBench.py).
> The script first creates indexes with 1M docs (based on
> SortableSingleDocSource, and based on wikipedia, if available).  Then
> it runs various combinations:
>   * Index with 20 balanced segments vs index with the "normal" log
>     segment size
>   * Queries with different numbers of hits (only for wikipedia index)
>   * Different top N
>   * Different sorts (by title, for wikipedia, and by random string,
>     random int, and country for the random index)
> For each test, 7 search rounds are run and the best QPS is kept.  The
> script runs singlePQ then multiPQ, and records the resulting best QPS
> for each and produces table (in Jira format) as output.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-dev-help@lucene.apache.org


Mime
View raw message