Mahesh Tailor wrote:
>Hello,
>
>TSM: 5.1.1.6
>OS: AIX 4.3.3
>Machine: IBM 6M1
>
>Hopefully this is a simple question: I have fourteen 36GB drives that
>are available for the diskpool and I was wondering whether it is better
>to have seven 5GB files or three 10GB files or one 35GB file or
>something else? The drives are mounted in two IBM-2014 Ultra-Wide SCSI
>disk drawers with separate Ultra-Wide contollers. The other 14 drives
>are used for DB, LOG, and spare.
>
>Thanks.
>
>Mahesh
>
>
Hi,
I have been following this thread and I think it is getting off track
from the original question. The original system in question was an AIX
box. However, several people have posted solutions relative to
Solaris. These are two entirely different beasts. Performance tuning
I/O for these 2 machines requires two completely different approaches.
In regards to AIX, I find that there is no performance advantage to
using "raw lv's" in this environment. Now to answer the question at
hand, since you have not provided us with the sizes of you DB and Log
nor the number of Random Access Storage pools that you require I will
give some general guidelines:
* DB Volumes; create several (3 to 15) DB volumes each on a
different file system(FS), have it consume the entire FS, each FS
on a separate physical volume(PV), use ITSM mirroring and make
sure the JFS log is on a separate PV. For additional performance
spread this across adapters as well.
* Log Volume ; Create one single volume that consumes an entire FS
on a single PV with no DB vols on it, locate the JFS log on a
separate PV. Use ITSM mirroring.
* JBOD Random Access Storage Pool Volumes; Create one FS on each PV,
one stg_pool volume per FS, put JFS logs on a separate disk,
spread across adapters if available.
* RAID 1 Random Access Storage Pool Volumes; Same as JBOD above just
use LVM mirroring. Make sure each copy is on separate PV and
don't bother mirroring the JFS log. Also, for newer versions of
AIX put this on a "Big VG" (mkvg -B) and use passive mirror write
consistency (mklv -w p). If this is not available on your
version, then disable mirror write consistency.
* RAID 10 (0+1) Random Access Storage Pool Volumes; Same as RAID 1,
but make sure you use plenty of PVs.
* RAID 5 Random Access Storage Pool Volumes; Create your arrays
with no more than 8 disk per array this will than become a single
hdisk, create as many arrays as needed. Create one vol for each
storage pool in a given array/hdisk. A storage pool can and
should use multiple arrays/hdisks if available. Make sure your
JFS log is located on a separate PV no need for it to be RAID 5.
* About JFS logs, I can't stress enough about separating JFS log
activity from the FS. In addition I prefer a separate log for
each FS, however they can be used by multiple FSs. But JFS log
write performance (and therefore FS performance) degrades at a
severe non-linear rate as increases in "In-Flight Transactions"
due to FS write activity occur, therefore separate logs will
minimize it.
* About VGs, I make no real mention of VGs here since it has no
direct relationship on performance, however keep in mind JFS logs
must be in the same VG as the FS. How you create your VGs is more
of an admin issue than performance.
* Performance tuning, there has been several discussion on this list
about tuning the AIX environment. There are huge performance
gains to be found in proper tuning. But there just isn't any
"cookbook" answer to tuning. You simple need to make the
measurements analyze the data and adjust accordingly. What makes
AIX so great is its ability to be highly tuned for a specific
task. Once you have done your measurement you can tune the kernel
with vmtune, adjust adapter and disk parms, or adjust some
application settings.
You will note that there are several options available for your "Random
Access Storage Pool Volumes" which one you chose is a function of the
hardware that is available to you. Clearly the best performing, most
available and of course most expensive option is to use RAID 10. I
think RAID 5 is the best middle ground solution for performance,
availability and cost. In regards to DB and Log volumes, I just haven't
found any better options than what is stated.
I did not attempt to justify any of my suggestions. My post has already
gone on quite long, however I would be willing to discuss any of these
items in greater detail if someone is interested.
--
Regards,
Mark D. Rodriguez
President MDR Consulting, Inc.
===============================================================================
MDR Consulting
The very best in Technical Training and Consulting.
IBM Advanced Business Partner
SAIR Linux and GNU Authorized Center for Education
IBM Certified Advanced Technical Expert, CATE
AIX Support and Performance Tuning, RS6000 SP, TSM/ADSM and Linux
Red Hat Certified Engineer, RHCE
===============================================================================
