On Mon, 26 Apr 2004, Rob Fielding wrote:
> scott.marlowe wrote:
> > On Fri, 23 Apr 2004, Chris Hoover wrote:
> >>DB's on Powervaults 220S using raid 5 (over 6 disks)
> > What controller is this, the adaptec? We've found it to be slower than
> > the LSI megaraid based controller, but YMMV.
> Wow, really? You got any more details of the chipset, mobo and kernel
> driver ?
We're running on a Dell 2650, the controller is the U320 LSI megaraid 2
channel (they only make the one that I know of right now). Don't know my
mobo chipset offhand, but might be able to find out what one dell includes
on the 2650. The kernel driver is the latest megaraid2 driver as of about
Feb this year.
> I've been taken to my wits end wrestling with an LSI MegaRAID 320-1
> controller on a supermicro board all weekend. I just couldn't get
> anything more than 10MB/sec out of it with megaraid driver v1 OR v2 in
> Linux 2.4.26, nor the version in 2.6.6-rc2. After 2 days of humming the
> Adaptec mantra I gave in and switched the array straight onto the
> onboard Adaptec 160 controller (same cable and everything). Software
> RAID 5 gets me over 40MB sec for a nominal cpu hit - more than 4 times
> what I could get out of the MegaRAID controller :( Even the 2nd SCSI-2
> channel gets 40MB/sec max (pg_xlog :)
> And HOW LONG does it take to detect drives during POST....ohhhh never
> mind ... I really just wanna rant :) There should be a free counseling
> service for enraged sysops.
I wonder if your controller is broken or something? Or maybe on a PCI
slow that has to share IRQs or something. I've had great luck with
SuperMicro mobos in the past (we're talking dual PPro 200 mobos, so
seriously, IN THE PAST here... ) Hell, my Dual PPro 200 with an old
MegaRAID 428 got 18 Megs a second cfer rate no problem.
Have you tried that lsi card in another machine / mobo combo? Can you
disable the onboard adaptec? We have on our Dell 2650s, the only active
controllers are the onboard IDE and the add in LSI-320-2 controller.
We're running ours with 128 Meg cache (I think could be 64) set to write
back. I think our throughput on a RAID-1 pair was somewhere around 40+
megs a second reads with bonnie++ With RAID-5 it was not really much
faster at reads (something like 60 megs a second) but was much more
scalable under heavy parellel read/write access for PostgreSQL.
Have you updated the BIOS on the mobo to see if that helps? I'm just
throwing darts at the wall here.
In response to
pgsql-performance by date
|Next:||From: Josh Berkus||Date: 2004-04-27 18:05:00|
|Subject: Re: Wierd context-switching issue on Xeon patch for 7.4.1|
|Previous:||From: Silke Trissl||Date: 2004-04-27 17:32:36|
|Subject: Join problem|