2. There definitely is a reason NOT to use RAID-5 for TSM. If you want
your TSM server to snake like a worm put DB and Log on RAID-5, especially
with a big stripe size - one of the best methods to shoot yourself in the
leg. RAID-5 for diskpool volumes also will have performance impact but not
so disastrous. It seems like you have enough disks to use mirroring or
RAID-1 so why not.
I still prefer to use raw LVs instead of filesystems despite the fact
community on this list cannot completely agree is there any significant
benefit or penalty for doing so. Using 'mklv -t tsm_db', '-t tsm_log' and
'-t tsm_vol' fairly well describes what those LVs are for and I am not
forced to exclude the files for the client on the server itself. IMO raw
LVs are few percent faster than JFS files, there is no need to worry about
large-enabled filesystems and file I/O buffering does not consume the
memory TSM server can use better.

1. 8x 9GB after mirroring will give you 36 GB. It looks to be enough for
both DB and Log. If your DB is not under 27 GB (cannot fit in three disks,
you wrote <30) and you cannot afford 9-th and 10-th disk (DB and Log on
separate disks gives best performance) you will be forced to use same
disks for DB+Log. Create 1-1.5 GB LV in the center of each disk for Log.
Create one or two LVs on the same disks for DB. It is recommended to
mirror them through TSM but you can also use AIX mirroring or RAID-1.

3. For both DB and diskpool volumes it is better to have one (IMO) volume
per disk. Other people on this list suggest more than one but opinions are
consolidated that more than few will have negative impact on performance.
Disk head moves will prevale disk reads/writes and performance will
suffer.

Zlatko Krastev
IT Consultant






"Varney, Patrick" <PAV5 AT PGE DOT COM>
Sent by: "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>
30.09.2002 22:21
Please respond to "ADSM: Dist Stor Manager"


        To:     ADSM-L AT VM.MARIST DOT EDU
        cc:
        Subject:        TSM Database Disk Layout Recommendations


Hello All,

I'm sure that this general question has been asked many times, but every
situation is different. And I can't seem to get to the www.adsm.org site,
so
here goes.

We just received some new SSA disks for our TSM servers and I'm looking
for
some advice on the best way to lay everything out. Here are the details on
our planned environment (the SSA configuration has not been done yet, so
it's flexible):

Server:
  RS/6000 7026-H70
  4-way processor
  4GB memory
  1 SSA HBA (2 loops)
  AIX 4.3.3 ML 10
  TSM 4.2.2.12
Internal SCSI Storage:
  4 x 9GB (current TSM database/log location)
SSA Loop 1:
  8 x 9GB
  RAID 5
  Approx. 60GB
SSA Loop 2:
  8 x 72GB
  RAID 5
  Approx. 500GB
TSM Database:  Less than 30GB
TSM Log: 3-4GB
Disk Storage Pools: 300GB

My current plan is to put the TSM database on the 9GB SSA drives (using
RAID
5, no TSM mirroring), the disk storage pool volumes on the 72GB SSA
drives,
and use the internal 9GB drives for database dumps to disk. However, I'm
open to suggestion. Specifically, I am looking for answers to the
following:

- Where is the best place for the TSM database log volumes? And if they
are
sharing space on the SSA disks, is there any advantage to creating a
separate file system (in general, is there any performance
advantage/disadvantage to multiple file systems on one SSA RAID array)?
- Is there any reason to NOT do RAID 5 on the SSA disks?
- For the database and disk storage volumes, is it better to create more
smaller (2GB) volumes or enable "large file" support and create fewer?

Any help would be greatly appreciated.

Thanks,

Patrick Varney
Pacific Gas & Electric Company
Fairfield Information Operations Center (FFIOC)
Enterprise Services  Group (ESG), Senior Network Specialist
        *  internal 8-227-2823
        *  external 707-436-2823
        *  pager 707-288-1513