From mboxrd@z Thu Jan 1 00:00:00 1970 From: Chris Mason Subject: Re: Chris: mason/patches/data-logging/experimental/2.6.1 upside down? Date: Tue, 13 Jan 2004 23:11:42 -0500 Message-ID: <1074053501.32707.995.camel@tiny.suse.com> References: <3171912.lEEDWjxvxS@spamfreemail.de> <200401132247.05606.Dieter.Nuetzel@hamburg.de> <31592854.1BIkuMAj52@spamfreemail.de> Mime-Version: 1.0 Content-Transfer-Encoding: quoted-printable Return-path: list-help: list-unsubscribe: list-post: Errors-To: flx@namesys.com In-Reply-To: <31592854.1BIkuMAj52@spamfreemail.de> List-Id: Content-Type: text/plain; charset="iso-8859-1" To: Jens Benecke Cc: reiserfs-list@namesys.com On Tue, 2004-01-13 at 17:13, Jens Benecke wrote: > Dieter N=C3=BCtzel wrote: >=20 > > See this, too. > > Re: v3 logging speedups for 2.6 > > It's (SuSE 9.0 2.6.1) is running fine with reiserfs-journal-writer. >=20 > I'll have a look. >=20 > What exactly do experimental/2.6.1/journal-writer and logging.gz do / > change / update? How experimental are they? >=20 journal-writer isn't experimental at all, it just removes old debugging code. reiserfs-logging.gz is the first half of the 2.4.x data logging code. Basically the 2.4.x data logging patch had two parts, the first part made small transactions significantly more efficient, and the second part implemented data=3Dordered/data=3Djournal. So, reiserfs-logging.gz has the small transaction performance improvements, along with some other rewrites of how log blocks and metadata get to disk. In 2.4.x and vanilla 2.6, reiserfs keeps log blocks and metadata pinned and then writes them out at very specific times. The new code keeps them pinned until they are safe to write, but lets the system pdflush daemons do some of the writeback. This makes for smoother io overall, since things aren't done in bursts. -chris