Can anybody point me in the right direction in regards to optimizing
performance on an IBM FastT 200 SAN?

We have our SAN up and running, but the performance is pitiful. We
can only get 12MBps throughput. Definitely not what we envisioned.

I've tweaked the numerous settings recommended by IBM in all their
optimization manuals, but can't get any faster throughput. We are
ready to chuck all this expensive equipment.

Any help would be greatly appreciated.

Tony

ajcinlv702@aol.com (Tony C.) wrote in message news:<b7b8cf61.0402152024.f145669@posting.google.c om>...
> Can anybody point me in the right direction in regards to optimizing
> performance on an IBM FastT 200 SAN?
>
> We have our SAN up and running, but the performance is pitiful. We
> can only get 12MBps throughput. Definitely not what we envisioned.
>
> I've tweaked the numerous settings recommended by IBM in all their
> optimization manuals, but can't get any faster throughput. We are
> ready to chuck all this expensive equipment.
>
> Any help would be greatly appreciated.
>
> Tony

What sort of support have you purchased ? who did you buy the fastt
from ? if you cannot get the correct support you have paid for then
raise a critsit via your account manager and get it escalated until
you get the answers you require.... your account manager should also
be able to provide initial setup support for your purchased
equipment.. dont throw it all away... kick some IBM account support
ass.... sounds like they need it..

Rgds
Mark

mky@talk21.com (mark taylor) wrote in message news:<dee0a7c3.0402160529.3520f143@posting.google. com>...
> ajcinlv702@aol.com (Tony C.) wrote in message news:<b7b8cf61.0402152024.f145669@posting.google.c om>...
> > Can anybody point me in the right direction in regards to optimizing
> > performance on an IBM FastT 200 SAN?
> >
> > We have our SAN up and running, but the performance is pitiful. We
> > can only get 12MBps throughput. Definitely not what we envisioned.
> >
> > I've tweaked the numerous settings recommended by IBM in all their
> > optimization manuals, but can't get any faster throughput. We are
> > ready to chuck all this expensive equipment.
> >
> > Any help would be greatly appreciated.
> >
> > Tony
>
> What sort of support have you purchased ? who did you buy the fastt
> from ? if you cannot get the correct support you have paid for then
> raise a critsit via your account manager and get it escalated until
> you get the answers you require.... your account manager should also
> be able to provide initial setup support for your purchased
> equipment.. dont throw it all away... kick some IBM account support
> ass.... sounds like they need it..
>
> Rgds
> Mark

We have the complete IBM support (hardware & software). And i"ve gone
round 'n' round with them. I've probably had 8 different support
calls that were left unresolved. Their main response is that the SAN
is 'functioning' (although not to it's promised speeds), and they have
to find an optimization expert on the SAN + W2K. And that is the last
I usually here from them.

My boss just started putting pressure on our hardware/software vendor
to facilitate this search.

Escalate, Escalate, Escalate!

Keep taking it to the next level if you don't get responses you like. When
Higher ups see unresolved calls with unhappy customers heads start to roll.

Also make sure you have the call at a high enough priority and they didn't
drop it back. If you performance is being affected badly then in my opinion
it is the same as a machine down.

Good Luck to you
Jeff

> We have the complete IBM support (hardware & software). And i"ve gone
> round 'n' round with them. I've probably had 8 different support
> calls that were left unresolved. Their main response is that the SAN
> is 'functioning' (although not to it's promised speeds), and they have
> to find an optimization expert on the SAN + W2K. And that is the last
> I usually here from them.
>
> My boss just started putting pressure on our hardware/software vendor
> to facilitate this search.

"Tony C." <ajcinlv702@aol.com> wrote in message
news:b7b8cf61.0402152024.f145669@posting.google.co m...
> Can anybody point me in the right direction in regards to optimizing
> performance on an IBM FastT 200 SAN?
>
> We have our SAN up and running, but the performance is pitiful. We
> can only get 12MBps throughput. Definitely not what we envisioned.
>
> I've tweaked the numerous settings recommended by IBM in all their
> optimization manuals, but can't get any faster throughput. We are
> ready to chuck all this expensive equipment.
>
> Any help would be greatly appreciated.
>
> Tony

Tony,

poor performance is an open arguement. FAStTs rely on appropriate RAID
configuration for the application. The usual I go for is 4+P with a segment
size of 64K and caching of 16K. If your I/O is more sequential read then you
can get away with higher number of disks in your array. If cache mirroring
is not important between controllers then that increase performance and so
does turning off caching for writes if the I/O is high on sequential type
writes. If it is random is nature then leave write caching on.

Your application should make good use of the block size settings of the RAID
array as well so if you have 64K for example in 4+P then your I/O block size
should be 256K. Does not need to be that high but multiples of 64K.

If you are using Windows then use IOMeter from microsoft to test best
configuration.

In any case a simple copy between folders is not going to give you a true
picture, because overall performance is obtained by having a number of
threads going at one time.

Regards Paresh

"Tony C." <ajcinlv702@aol.com> wrote in message
news:b7b8cf61.0402152024.f145669@posting.google.co m...
> Can anybody point me in the right direction in regards to optimizing
> performance on an IBM FastT 200 SAN?
>
> We have our SAN up and running, but the performance is pitiful. We
> can only get 12MBps throughput. Definitely not what we envisioned.
>
> I've tweaked the numerous settings recommended by IBM in all their
> optimization manuals, but can't get any faster throughput. We are
> ready to chuck all this expensive equipment.
>
> Any help would be greatly appreciated.

Nobody else seems to have chimed in with real performance figures for the
FAStT200, but our experience is the same as Tony's.
With an 8 disk RAID5 array on a p630 (dual pathed), no combination of
segment size, cache settings, disk queue depth or filesystem
type could get us much more than Tony reported. I think we got to the heady
heights of ~17MB/s with a JFS2/inline log and a
segment size of 64k. We tried the latest firmware and SM8.3 without
improvement. If anybody is doing substantially better, I'd love
to hear how (AIX5.1ML05). The response I got from our IBM Storage Guy was
"people who care about performance don't
normally buy the 200"...

I agree... escalate, raise a critsit stating you are totally unhappy
with the support you have paid for. IBM Support is driven by customer
feedback.. if they think they are doing ok they are highly unlikely to
change... make them change...

ajcinlv702@aol.com (Tony C.) wrote in message news:<b7b8cf61.0402152024.f145669@posting.google.c om>...
> Can anybody point me in the right direction in regards to optimizing
> performance on an IBM FastT 200 SAN?
>
> We have our SAN up and running, but the performance is pitiful. We
> can only get 12MBps throughput. Definitely not what we envisioned.
>
> I've tweaked the numerous settings recommended by IBM in all their
> optimization manuals, but can't get any faster throughput. We are
> ready to chuck all this expensive equipment.
>
> Any help would be greatly appreciated.
>
> Tony

My experience isn't all that great as far as performance. We got it
"acceptable". Ours are strictly Oracle DB boxes, p650s, connected to
the FAStT 200 SAN (using a switch). After much tweaking, the biggest
gain was when the IBM'er talked to the dba about his settings, and
found that the sort_area_size parameter was at the default 64k, which
when raised to 1 or 2mb made a world of difference. Not much has been
posted in this newsgroup as far as specific settings (I know, a while
back I was looking), but I'm sure that's because the different
environments need to be handled individually. I'll post some of my
specifics, in case it will help you at all, but this comes with a
*big* caveat. The previous posters who mention escalating and
demanding help from support, as well as realizing the limitations, are
still right on the money. Since our servers have differing amounts of
processor and memory, each with some different settings, I'll pick one
for an example:

p650, two 1.2ghz procs, 6gb RAM, AIX 5.2, fibre card is fc6228 (2gb)

Fibre-channel adapter
---------------------
num_cmd_elems: 2048
lg_term_dma: 0x1000000
max_xfer_size: 0x1000000

Memory I/O Settings
-------------------
minfree: 240
maxfree: 368
maxperm%: 40
minperm%: 10
maxclient%: 40
strict_maxperm: 1
maxpgahead: 128
maxrandwrt: 128
numclust: 16
sync_release_ilock: 1

Hdisk Attributes
----------------
prefetch_mult: 8
queue_depth: 40

Asynchronous I/O
----------------
maxservers: 20
minservers: 10
maxrequests: 8192

Storage Manager 8.3
-------------------
Cache block size: 16k
Segment size: 64k
Modification Priority: Medium
Flush/Fill % for cache: 22/20
caching (per LUN): read: off, write: on, write w/ mirror: off

Oracle User Limits increased
----------------------------
data: 1024000
rss: 1024000
stack: 1024000


Regards,
Steve