With only 6 nodes pushing data to the TSM server at a time, what kind of
throughput are you expecting the TSM server to have to handle (is the
network going to limit throughput? How much data do you expect the nodes
themselves to push?) I.e. do you really think with that many disks in your
disk pool you're going to see some discernable difference between jfs and
raw logical volumes? You'll see terrific performance regardless of which
route you go but you need to leverage the advantage of having so many
disks...
Keep in mind how TSM uses disk storage pool volumes. In your below scenario
you suggest having 32 disk pool volumes and yet you also mention only appox
6 nodes will be backing up at once. Assuming only 1 thread per node, you'll
only be accessing 6 of those 32 disk volumes at any one time. The other's
will sit idle while each node essentially backs up to only 1 disk at a time.
You wouldn't be leveraging the advantage of having so many disks..
Using striping or even RAID-5 makes much more sense. E.g. you could do as
Rich mentioned below. Or perhaps 6 RAID-5 arrays of 5 disks each with 2 hot
spares? Certainly makes admining easier since you'll never lose data. You're
going to have to take all these variables into consideration. How is TSM
going to hit those volumes you create? How is AIX going to write to the
disks given the configuration you're considering? How are the SSA adapters
and pathways going to handle the given configuration?
Gerald Wichmann
Sr. Systems Development Engineer
Zantaz, Inc.
925.598.3099 w
408.836.9062 c
-Original Message-
From: Andrew Carlson [mailto:[EMAIL PROTECTED]]
Sent: Thursday, May 09, 2002 11:39 AM
To: [EMAIL PROTECTED]
Subject: Re: comments on disk storage plan
I haven't seen the same thing as you. I was using raw logical volumes,
and converted to JFS. I saw little if no penalty during write, but a
big win during the sequential reads. As always, your mileage may vary,
but I would not rule out JFS if I was you.
Andy Carlson|\ _,,,---,,_
Senior Technical Specialist ZZZzz /,`.-'`'-. ;-;;,_
BJC Health Care|,4- ) )-,_. ,\ ( `'-'
St. Louis, Missouri '---''(_/--' `-'\_)
Cat Pics: http://andyc.dyndns.org/animal.html
On Thu, 9 May 2002, Rich Brohl wrote:
Chris,
I would create a stripped logical volume using as many disk as possible
(the more the better) in the Volume Group. Also, do not create a
filesystem just leave the Logical Volume in the Volume Group. This will
keep you from having to deal with the AIX JFS overhead. When you define
your volumes to the disk storage pool point it to the (r)logical volume
name. For instance if you create a logical volume in the volume group
named tsmstg1lv, then in TSM define you volume /dev/rstmstg1lv to the
disk
storage pool. This will give you some awesome throughput.
On the flip side when you do your migration to tape you will be doing
mostly sequential reads from the disk storage pools so the can keep the
tape drive(s) running at max write speed.
Rich Brohl
ISM Support
Tie Line 547-9317
Direct 303-693-4969
Pager 888-524-9030
chris rees [EMAIL PROTECTED]@VM.MARIST.EDU on 05/09/2002 08:46:02
AM
Please respond to ADSM: Dist Stor Manager [EMAIL PROTECTED]
Sent by:ADSM: Dist Stor Manager [EMAIL PROTECTED]
To:[EMAIL PROTECTED]
cc:
Subject:comments on disk storage plan
Hi all,
Just wanted to sanity check my disk storage pool layout I'm planning.
AIX 4.3.3
Server H80
2 x D40 SSA drawers, two loops over two SSA 160 adapters
32 x 9Gb SSA disks
Max nodes backing up at any one time is approx 6
I don't want to mirror the volumes
32 x 9Gb logical volumes, each logical volume on one SSA disk. These lvs
will then translate into 32 disk storage pool volumes. This way if I have
a
disk failure I only lose one disk pool volume.
Anyone got any ideas what kind of read/write performance I should expect
to
see from this config?
Any other options considered.
Regards
Chris
_
Get your FREE download of MSN Explorer at
http://explorer.msn.com/intl.asp.