public inbox for linux-xfs@vger.kernel.org
 help / color / mirror / Atom feed
From: Bill Kendall <wkendall@sgi.com>
To: xfs@oss.sgi.com
Subject: [PATCH v2 0/8] xfsdump: enable support for multiple streams
Date: Mon,  7 Nov 2011 14:58:23 -0600	[thread overview]
Message-ID: <1320699511-12281-1-git-send-email-wkendall@sgi.com> (raw)

Version 2 of this series changes only patch #3. POSIX semaphores
are now used instead of rolling our own, as suggested by Christoph.
We lose the ability to count the number of waiting threads, which
was used in some asserts. There was adequate coverage in related
asserts, so this was deemed acceptable.


This series resurrects the IRIX multi-stream support for splitting a
backup among several output files/tapes. This offers some nice
performance improvements, particularly in xfsrestore where a single
stream often cannot keep the filesystem/disks busy. I've observed
a 1.7x improvement on a backup and a 5x improvement on restore.

I have a couple of xfstests for this, and will submit those once
a few outstanding xfsdump test patches have been reviewed.

There's a bit more work to do:

  - Now that xfsdump has threads once again, the tape I/O ring
    buffer support can be enabled. This series leaves it disabled
    so that more testing can be done in that area.

  - Currently the stream split points are determined by doing an
    extra inode scan. This is unchanged from how it was done on
    IRIX. I'd like to change this so that enough info is kept from
    the initial inode scan to determine appropriate split points
    without an additional scan.

  - You may notice in the last patch of this series that a lot
    of "miniroot" references also look at whether or not the dump
    is to a pipe. Now that the "miniroot" checks are gone, it's
    possible to clean up the pipe-related code too. I've got a
    separate patch series for that which I'll submit later.

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

             reply	other threads:[~2011-11-07 20:59 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2011-11-07 20:58 Bill Kendall [this message]
2011-11-07 20:58 ` [PATCH v2 1/8] xfsdump: link with libpthread Bill Kendall
2011-11-08  2:02   ` Alex Elder
2011-11-07 20:58 ` [PATCH v2 2/8] xfsdump: remove multi-stream synchronous dir dump Bill Kendall
2011-11-08  2:02   ` Alex Elder
2011-11-07 20:58 ` [PATCH v2 3/8] xfsdump: implement lock abstraction with pthreads Bill Kendall
2011-11-08  2:02   ` Alex Elder
2011-11-07 20:58 ` [PATCH v2 4/8] xfsdump: simplify qlock ordinal bitmap Bill Kendall
2011-11-08  2:02   ` Alex Elder
2011-11-07 20:58 ` [PATCH v2 5/8] xfsdump: convert IRIX sproc threads to pthreads Bill Kendall
2011-11-08  2:02   ` Alex Elder
2011-11-07 20:58 ` [PATCH v2 6/8] xfsdump: process thread exit status Bill Kendall
2011-11-08  2:03   ` Alex Elder
2011-11-07 20:58 ` [PATCH v2 7/8] xfsdump: path lookup cache must be thread specific Bill Kendall
2011-11-08  2:03   ` Alex Elder
2011-11-07 20:58 ` [PATCH v2 8/8] xfsdump: enable multiple streams Bill Kendall
2011-11-08  2:03   ` Alex Elder
2011-11-10 11:00 ` [PATCH v2 0/8] xfsdump: enable support for " Christoph Hellwig
2011-11-10 16:22   ` Bill Kendall

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1320699511-12281-1-git-send-email-wkendall@sgi.com \
    --to=wkendall@sgi.com \
    --cc=xfs@oss.sgi.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox