[
https://issues.apache.org/jira/browse/LUCENE-2324?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12988082#action_12988082
]
Michael McCandless commented on LUCENE-2324:
--------------------------------------------
bq. Where are we gaining flush concurrency then?
Even if we do full stop for commit and getReader, we've still gained
concurrency on the batch indexing case.
So, I think we can actually avoid full stop for commit/getReader.
The current plan w/ deletes is that a delete gets buffered 1) into the global
pool (stored in DW and pushed whenever any DWPT flushes), as well as 2) per
DWPT. The per-DWPT pools apply *only* to the segment flushed from that DWPT,
while the global pool applies during coalescing (ie to all "prior" segments).
To avoid the full-stop, I think during the flush we can have two global delete
pools. We carefully sweep all DWPTs and flush each, in succession. Any DWPT
not yet flushed is free to continue indexing as normal, putting deletes into
the first global pool, flushing as normal. But, a DWPT that has been flushed
by the "sweeper" must instead put deletes for an updateDocument carefully into
the 2nd pool, and not buffer the delete into DWPTs not yet flushed.
Basically, as the sweeper visits each DWPT, it's segregating them into "pre
commit point" and "post commit point".
We also must then ensure that no "post commit point" DWPT is allowed to flush
until all "pre commit point" DWPTs have been visited by the sweeper.
> 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
