Date: Sat, 19 Jan 2002 14:22:57 -0800 From: Kent Stewart <kstewart@owt.com> To: Wilko Bulte <wkb@freebie.xs4all.nl> Cc: Murray Stokely <murray@FreeBSD.ORG>, qa@FreeBSD.ORG Subject: Re: RELENG_4_5 Branch : January 21 ? Message-ID: <3C49F1C1.8010408@owt.com> References: <20020119015249.GD18200@freebsdmall.com> <20020119084007.A10312@freebie.xs4all.nl> <20020119111145.GL18200@freebsdmall.com> <20020119190247.A12050@freebie.xs4all.nl> <3C49EA37.1090601@owt.com> <20020119230428.A39931@freebie.xs4all.nl>
next in thread | previous in thread | raw e-mail | index | archive | help
Wilko Bulte wrote: > On Sat, Jan 19, 2002 at 01:50:47PM -0800, Kent Stewart wrote: > > >>Wilko Bulte wrote: >> >> >>>On Sat, Jan 19, 2002 at 03:11:45AM -0800, Murray Stokely wrote: >>> >>> >>>>On Sat, Jan 19, 2002 at 08:40:07AM +0100, Wilko Bulte wrote: >>>> > > .. > > >>>> I put a new 7200RPM disk into my PWS500 and got my make world time >>>>down to just over 3 hours. As an exercise in patience I've just >>>>started a "make release" to see how long the build will take. Is disk >>>> >>>> >>>IIRC it something like 6 hours or so on my DS10 >>> >>> >>> >>>>I/O this slow on all FreeBSD/Alpha machines, or is it just mine? >>>>bonnie++ results are pathetic compared to any x86 box on my network. >>>> >>>> >>>IDE or SCSI disk? I have not tested/compared IOspeed to be honest. >>> >>> >>> >> >>I was testing some ATA-66/100 systems and I hit a limit doing >>buildworlds. The build speed became much faster when I used 2 or more >>HDs. It was really important to get /usr/src and /usr/obj and >>different HDs. I was dropping buildworld times on a dual 866 >>coppermines from something like 40 minutes to 26 minutes by doing >>this. I ended up thinking about U2W but decided not to spend the >>money. The 3 ATA-100 drives cost what a single 30GB lvd scsi cost. >> >>I log all of my builds and eventually ended up with the log and system >>on one drive and /usr/src and /usr/obj on the other two. I figured >>splitting things up would also benefit a scsi system because it would >>take advantage of more of the bandwidth from the controller to the HDs. >> > > As far as I can see on Alpha we are maxing out the CPU instead of the disks. > I have UW SCSI on my DS10, on a Qlogic 1040. Not the newest of drives > though. I think the CVS operation is disk bound, my CVS repo is on NFS > via 100 mbit ethernet to worsen things. I was moving up in cpu speed and around 800 MHz on the Intel/AMD cpu's the HD's appeared to be the bottleneck. The cpu appeared to be the bottleneck below that speed. One thing I forgot to mention was each IDE HD has its own controller. I was running "time mkworld". Mkworld was a local logging script and I was looking at cpu % and elapsed wall clock time. I am using a 1600+ XP right now and seeing < 80% on the cpu. Setiathome gets the rest of the time but not running setiathome didn't change anything. Specifying a -j? on a single cpu also increased the elapsed build times in most of the configurations. Kent -- Kent Stewart Richland, WA mailto:kbstew99@hotmail.com http://users.owt.com/kstewart/index.html FreeBSD News http://daily.daemonnews.org/ To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-qa" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3C49F1C1.8010408>