Miles' remarks are good with one exception - you do not need to lose
availability to get better performance.
The SSA drawer allows you to allocate disks to loops in groups of four,
i.e. not to have the whole drawer with only left and right ends.
So you can split evenly the loop:
/--- disks 1-4 --- disks 5-8 ---\
ssa0 portA ssa1 portA
\--- disks 13-16 --- disks 9-12 ---/
Connect second drawer with same topology to ports B of ssa0 & ssa1. Play
around with ssaraid tool and allocate disks 1-4 & 13-16 to have adapter
ssa0 as primary and disks 5-8 & 9-12 to have ssa1 as primary. Thus without
losing adapter redundancy you gain the performance of four loops scenario.
If the chain is something like ssa0 -- ssa1 -- whole_drawer you are
sacrificing two SSA concurrent paths thus losing parallelism (and
corresponding performance).
This is another wheel I did not invented - IBM ESS employs such loop
topology to get maximum performance. Next trick is to define disk order
appropriately for exploit those parallel paths (i.e. define striping over
disks 1, 9, 5, 13, 2, 10, etc.). Would this be done through AIX LVM
striping or through TSM DB, log & disk volumes does count but not so much
as proper SSA routing.
For more details look at a nice redbook "Understanding SSA Subsystems in
Your Environment" (SG24-5750-00) and read thoroughly section 3.10 "How to
develop an SSA configuration".
Zlatko Krastev
IT Consultant
P.S. I guess you already know but to mention anyway that any port of ssa0
can be on same loop with any port of ssa1 but you should not cable ports A
& B of the same adapter within a loop.
Zlatko
Please respond to "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>
Sent by: "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>
To: ADSM-L AT VM.MARIST DOT EDU
cc:
Subject: Re: Disk Availability
Hi Goeff,
there are many things that you can do, depending on what you want.
However, I would not be so concerned about an adapter failing, failing
disks are much more common.
Here are some things to do/consider:
1. put each adapter on a separate PCI bus.
2. balance the disks over the adapters.
3. if you really don't care if a disk fails, use RAID 0, if the system
can't go down use RAID 5.
4. Lately I have been using hardware RAID and LVM striping, the
reliability and performance is very good.
5. If you are putting you TSM DB and LOG on these disks: I might not use
any kinda of striping: JBOD or RAID 1, is probably better. Using TSM
mirroring across SSA adapter is probably best too.
6. if your using filesystems on these disks, use RAID and LVM striping.
Miles
----------------------------------------------------------------------------------------------
Miles Purdy
Miles Purdy
System Manager
Farm Income Programs Directorate
Winnipeg, MB, CA
purdym AT fipd.gc DOT ca
ph: (204) 984-1602 fax: (204) 983-7557
"If you hold a UNIX shell up to your ear, can you hear the C?"
-------------------------------------------------------------------------------------------------
>>> GEOFFREY.L.GILL AT SAIC DOT COM 29-Mar-02 5:53:42 PM >>>
>>> GEOFFREY.L.GILL AT SAIC DOT COM 29-Mar-02 5:53:42 PM >>>
Hi all,
I'm working on installing a second SSA drawer in my system. I'd like any
opinions as to best connect this. What I've got is an M80 with 2 SSA
controllers, and now, 2 SSA drawers full of 36gb disks. What would be the
best way to connect this so no matter what happens to the controllers all
the disk would still be available? Is there a different solution I should
use?
Thanks for the help.
Geoff Gill
TSM Administrator
NT Systems Support Engineer
SAIC
E-Mail: <mailto:gillg AT saic DOT com> gillg AT saic DOT com
Phone: (858) 826-4062
Pager: (877) 905-7154
|
