Re: SW RAID read performance

2006-08-08 Thread Matus UHLAR - fantomas
On 07.08.06 12:51, David Siroky wrote: > So what about Linux implementation? I think linux RAID driver can benefit from having data stored on multiple places in RAID1... > > On Fri, 04 Aug 2006, Justin Piszcz wrote: > > > AFAIK RAID1 offers faster seeks to the data, but not striped reads. > Henr

Re: SW RAID read performance

2006-08-07 Thread David Siroky
So what about Linux implementation? Henrique de Moraes Holschuh píše v Pá 04. 08. 2006 v 11:30 -0300: > On Fri, 04 Aug 2006, Justin Piszcz wrote: > > AFAIK RAID1 offers faster seeks to the data, but not striped reads. > > RAID1 offers, in theory, stripped reads with an arbitrary stripe size. It

Re: SW RAID read performance

2006-08-04 Thread Henrique de Moraes Holschuh
On Fri, 04 Aug 2006, Justin Piszcz wrote: > AFAIK RAID1 offers faster seeks to the data, but not striped reads. RAID1 offers, in theory, stripped reads with an arbitrary stripe size. It depends only on the intelligence of the RAID1 implementation. -- "One disk to rule them all, One disk to fi

Re: SW RAID read performance

2006-08-04 Thread Justin Piszcz
On Fri, 4 Aug 2006, David Siroky wrote: Dave Ewart píše v Čt 03. 08. 2006 v 19:31 +0100: On Thursday, 03.08.2006 at 07:30 -0400, Justin Piszcz wrote: I have 2 disks in RAID 1 and I expected reading will be approx. 2x faster then a single disk reading because there can be used the same techn

Re: SW RAID read performance

2006-08-04 Thread David Siroky
Dave Ewart píše v Čt 03. 08. 2006 v 19:31 +0100: > On Thursday, 03.08.2006 at 07:30 -0400, Justin Piszcz wrote: > > > >I have 2 disks in RAID 1 and I expected reading will be approx. 2x > > >faster then a single disk reading because there can be used the same > > >technique as on RAID 0. I tested

Re: SW RAID read performance

2006-08-03 Thread Justin Piszcz
Hm ah nice. avg-cpu: %user %nice %system %iowait %steal %idle 0.000.005.05 94.950.000.00 Device:rrqm/s wrqm/s r/s w/srkB/swkB/s avgrq-sz avgqu-sz await svctm %util sda277.78 0.00 100.00 1.01 23660.61 4.04 468.56 0.68 6.7

Re: SW RAID read performance

2006-08-03 Thread Dave Ewart
On Thursday, 03.08.2006 at 15:30 -0400, Justin Piszcz wrote: > I've only ran the benchmarks on a single controller. Run a test for me, > read a lot of data on the drive and show us iostat -x -k 1 and see if it > reads from both drives at the same time. http://www.sungate.co.uk/tmp/iostat.log -

Re: SW RAID read performance

2006-08-03 Thread Justin Piszcz
I've only ran the benchmarks on a single controller. Run a test for me, read a lot of data on the drive and show us iostat -x -k 1 and see if it reads from both drives at the same time. On Thu, 3 Aug 2006, Dave Ewart wrote: On Thursday, 03.08.2006 at 07:30 -0400, Justin Piszcz wrote: I hav

Re: SW RAID read performance

2006-08-03 Thread Owen Heisler
On 8/3/06, Dave Ewart <[EMAIL PROTECTED]> wrote: On Thursday, 03.08.2006 at 07:30 -0400, Justin Piszcz wrote: > From my benchmarking on my RAID1 (dual raptor RAID1) it appears it > reads from one drive or the other. It all depends on which disk controllers you're using and their throughput. I h

Re: SW RAID read performance

2006-08-03 Thread Dave Ewart
On Thursday, 03.08.2006 at 07:30 -0400, Justin Piszcz wrote: > >I have 2 disks in RAID 1 and I expected reading will be approx. 2x > >faster then a single disk reading because there can be used the same > >technique as on RAID 0. I tested it with bonnie++. Can linux SW RAID > >1 read "in stripes"

Re: SW RAID read performance

2006-08-03 Thread Justin Piszcz
On Thu, 3 Aug 2006, David Siroky wrote: Hi! I have 2 disks in RAID 1 and I expected reading will be approx. 2x faster then a single disk reading because there can be used the same technique as on RAID 0. I tested it with bonnie++. Can linux SW RAID 1 read "in stripes" to boost the reading per