[
https://issues.apache.org/jira/browse/LUCENE-2324?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12983246#action_12983246
]
Michael Busch commented on LUCENE-2324:
---------------------------------------
{quote}
I ran a quick perf test here: I built the 10M Wikipedia index,
Standard codec, using 6 threads. Trunk took 541.6 sec; RT took 518.2
sec (only a bit faster), but the test wasn't really fair because it
flushed @ docCount=12870.
{quote}
Thanks for running the tests!
Hmm that's a bit disappointing - we were hoping for more speedup.
Flushing by docCount is currently per DWPT, so every initial segment
in your test had 12870 docs. I guess there's a lot of merging happening.
Maybe you could rerun with higher docCount?
bq. But I can't test flush by RAM - that's not working yet on RT right?
True. I'm going to add that soonish. There's one thread-safety bug
related to deletes that needs to be fixed too.
{quote}
Then I ran a single-threaded test. Trunk took 1097.1 sec and RT took
1040.5 sec - a bit faster! Presumably in the noise (we don't expect
a speedup?), but excellent that it's not slower...
{quote}
Yeah I didn't expect much speedup - cool! :) Maybe because some
code is gone, like the WaitQueue, not sure how much overhead that
added in the single-threaded case.
{quote}
I think we lost infoStream output on the details of flushing? I can't
see when which DWPTs are flushing...
{quote}
Oh yeah, good point, I'll add some infoStream messages to DWPT!
> Per thread DocumentsWriters that write their own private segments
> -----------------------------------------------------------------
>
> Key: LUCENE-2324
> URL: https://issues.apache.org/jira/browse/LUCENE-2324
> Project: Lucene - Java
> Issue Type: Improvement
> Components: Index
> Reporter: Michael Busch
> Assignee: Michael Busch
> Priority: Minor
> Fix For: Realtime Branch
>
> Attachments: LUCENE-2324-SMALL.patch, LUCENE-2324-SMALL.patch,
> LUCENE-2324-SMALL.patch, LUCENE-2324-SMALL.patch, LUCENE-2324-SMALL.patch,
> LUCENE-2324.patch, LUCENE-2324.patch, LUCENE-2324.patch, lucene-2324.patch,
> lucene-2324.patch, LUCENE-2324.patch, test.out, test.out, test.out, test.out
>
>
> See LUCENE-2293 for motivation and more details.
> I'm copying here Mike's summary he posted on 2293:
> Change the approach for how we buffer in RAM to a more isolated
> approach, whereby IW has N fully independent RAM segments
> in-process and when a doc needs to be indexed it's added to one of
> them. Each segment would also write its own doc stores and
> "normal" segment merging (not the inefficient merge we now do on
> flush) would merge them. This should be a good simplification in
> the chain (eg maybe we can remove the *PerThread classes). The
> segments can flush independently, letting us make much better
> concurrent use of IO & CPU.
--
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: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
