I will not disagree that there may be a better way to configure our disks, I
was trying to point out one way to determine how to configure the disk pools or
determine that the disk pools are a problem.
We have had many talks with EMC, they said early on that the RAID group size
was optimized for a particular number of disks as you described. When we went
to Flare 12 they said that was no longer true. It is hard to tell when the
information we get from EMC is correct or not.
In our case, the disks are not a bottleneck. We run our trunked 2Gb Ethernet
at or near 75% and we see over 20MB/sec on our 3590-E tape drives.
On your ATA drives, how do you have your storage pools broken up? What size
are the STG volumes? What size are the LUNs? 320 or 250GB disks? We will be
upgrading all of the backend parts to our TSM servers this summer.
Andy Huebner
-----Original Message-----
From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU] On Behalf
Of Tab Trepagnier
Sent: Saturday, May 14, 2005 3:48 PM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: Re: [ADSM-L] Diskpool question
Another factor to consider on EMC is the RAID type. According to their
white paper, "EMC Backup Storage Solutions: Backup-to-Disk Guide with IBM
Tivoli Storage Manager", the preferrred RAID type is **_RAID-3_**. Our
Clariion requires five or nine disks to implement that RAID level. With a
pod full of SATA disks we implemented two five-disk RAID-3 arrays and one
four-disk RAID-5 array for use as TSM disk pools. We consistently see
40-50 MB/s with the RAID-3 LUNs, while the RAID-5 LUN seems to max out at
about 22 MB/s. I understand that drive count is a factor in all this, but
I doubt that the difference of one disk's media would yield a 50% loss of
performance. However, in subsequent discussions with their engineers
they've stated that the Clariion internal "Flare" code is optimized for
RAID groups made with (4 x n + 1) disks; meaning 5, 9, 13, etc. Using any
other drive count in a RAID group compromises that optimization. At some
point, we will probably remake some of our RAID groups, including that one
RAID-5 array.
Tab Trepagnier
TSM Administrator
Laitram, L.L.C.
"ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU> wrote on 05/11/2005
01:54:04 PM:
> Andy Huebner wrote:
>
> >We use RAID 5 (5+1) on a Clariion using 146GB drives. We peak the
> drives at over 55MB/sec, sustained speed is over 45MB/sec. How fast
> are you reading and writing to your disks? If it is faster than
> your SAN storage can read and write from RAID 5 then mirroring or
> striping is a better choice, although I would not use unprotected disks.
> >
> >Andy Huebner
> >
> > -----Original Message-----
> >From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU] On
> Behalf Of Luc Beaudoin
> >Sent: Tuesday, May 10, 2005 7:47 AM
> >To: ADSM-L AT VM.MARIST DOT EDU
> >Subject: [ADSM-L] Diskpool question
> >
> >My diskpool is on a SAN, I have to increase the size cause I'm taking
> >about 450GB per day and my diskpool is 350GB...
> >
> >Is there a RAID issue for performance .... right now my diskpool is on
a
> >RAID5 ... should I still use RAID5 or there a better RAID
configuration??
> >
> >thanks
> >
> >Luc
> >
> >
> >This e-mail (including any attachments) is confidential and may be
> legally privileged. If you are not an intended recipient or an
> authorized representative of an intended recipient, you are
> prohibited from using, copying or distributing the information in
> this e-mail or its attachments. If you have received this e-mail in
> error, please notify the sender immediately by return e-mail and
> delete all copies of this message and any attachments.
> >Thank you.
> >
> >
> >
> >
> Just for information, EMC did not optimize the FlareCode on Clariion for
> RAIDGroup with even disk (3+1, 5+1, ...) so It could be better to use
> 4+1, 6+1, 8+1 RAIDGroup (See Clariion Best Practises on Powerlink). We
> use a 4+1 configuration and during our backup database we have a
> sustained speed of 70-80 MB/s.
>
> Stripe on each disk is 64KB lenght, for four disk you have a full data
> stripe of 256KB (4x64KB + 64KB Parity) and th default LTG size of AIX
> Volume Group is 128kB.
|
