[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: wc_db performance (was: wc_db API discussion)

From: Branko Čibej <brane_at_e-reka.si>
Date: Sat, 12 Mar 2011 02:30:03 +0100

On 12.03.2011 01:11, Mark Phippard wrote:
> I am glad you sent this because I was getting ready to send an email
> to see if anyone is looking into the suggestions you have made here.
> I think we have to get this work done soon. We cannot release with
> performance like it is. How do we define the scope of the work that
> needs to be done so that we can divide and conquer and get these
> changes in place?

From the various performance-related threads, I can extract two major
areas of improvement:

    * transactional writing of changes to the wc-db (especially during
      checkout/update), where database updates are currently done via
      many tiny transactions.
    * reducing the number of queries performed during wc read/scan
      operations.

The first task doesn't look at all trivial from where I'm standing. I
admit to being blissfully ignorant of how checkout/update twiddle the wc-db.

For the second task, I think the first order of business is to change
the wc-db tree crawler to do one query instead of zillions, or at least,
where several queries are required, to do them all in one transaction.
The first step requires "someone" to catalog all the places where it's
used, and how the results are being filtered by the callers; based on
that, we can come up with a reasonably small set of parameters that
would make that crawling query useful for all the myriad uses the
crawler currently has. This part is more or less linear; afterwards,
callers can probably be updated as needed.

Then there's the NODES/ACTUAL_NODE merge. That one is a bit tricky, but
I think I can see a way to create (and use) such a merged table whilst
still keeping the current separate ones in use. The problem is that
doing this would essentially double the size of the wc-db, *but* it
would allow a gradual transition to the merged table, and the transition
would not have to be complete before the 1.7 release.

(There's also a "proof-of-concept" path that requires creating a view
that simulates such a merged table, but it has two serious drawbacks:
(1) a view is read-only, so data modifications could only happen through
the original tables; and (2) the query for such a view would be
prohibitively slow so there'd be no way to keep track of possible
performance improvements.)

-- Brane
Received on 2011-03-12 02:30:38 CET

This is an archived mail posted to the Subversion Dev mailing list.