Date: Thu, 28 Aug 2003 18:22:17 -0700 From: David Schultz <das@FreeBSD.ORG> To: Claus Guttesen <cguttesen@yahoo.dk> Cc: Bill Moran <wmoran@potentialtech.com> Subject: Re: Some additional tests run on my performance testing Message-ID: <20030829012217.GA47242@HAL9000.homeunix.com> In-Reply-To: <20030828204032.97741.qmail@web14102.mail.yahoo.com> References: <20030828171955.GE83759@perrin.nxad.com> <20030828204032.97741.qmail@web14102.mail.yahoo.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Aug 28, 2003, Claus Guttesen wrote: > So I newfs'ed it with 8 and 16 kb blocksize did an > import of a 1.5 GB pg-dump. > > > numbers you suggest above, I loaded a DB with 8k and > > 16K blocks > > (translation: almost all write activities). > > > > them to stay about the same across the board. If > > someone wants to do > > some good read tests, I'd be interested in those > > results. > > > > The 8 kb blocksize took 60 min. to import, and the 16 > kb ditto took 45 min. So I'm settling on 16 kb blocks. Note that doing sequential reads and writes is VERY different from doing random reads and writes. In the former case, a larger filesystem block size may reduce fragmentation and help you, but in the latter case---specifically for writes---you're doing twice the I/O due to the larger filesystem blocks. Your import and Bill's tests fall into the first category. It would be interesting to see what sort of performance you get under a real workload.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030829012217.GA47242>