On Wed, 21 Oct 1998 15:28:40 -0400, Bill Lancaster <[log in to unmask]>
wrote:

>At 04:21 AM 10/21/98 GMT, David A. Lethe wrote:
>>On Tue, 20 Oct 1998 13:39:24 -0400, Forrest Smith
><snip>
>>Several things:
>>
>>1. HP WILL support customers with SSA attached to both HP3Ks and
>>HP9Ks, despite what some people have posted (they did, however,
>>qualify that their information was old). If you need names & numbers,
>>then contact me offline.
>>
>
>Sorry, David, but isn't this a bit obfuscatory?  This issue I have raised
>in the past isn't that HP won't support the *customer* but that HP
>(currently) won't support the *product*.  I'm still holding out for
>official product support by HP even if they simply say "SSA is OK.  Use it
>with our blessing."

HP will basically do the same thing it does with any non-HP product.
i.e., imagine a customer has non-HP RAM.  In the event there is a
problem then HP will run whatever diagnostics are required to isolate
the problem.

In the event the diagnostics reveal the problem is with non-HP gear,
then HP will not replace or even touch it.

Furthermore, at the discresion of the regional CE mgr, HP may decide
to put the non-HP gear on a hardware contract.  I know of one customer
where HP put a bid out to put the SSA on full support.  (HP lost the
bid, but they tried to get the business!)

HP has not officially said "SSA is OK", however, they did spend a
solid month testing it, and identified a few problems which were
repaired with HP's help!  I believe the reason why HP hasn't embraced
SSA is because of pressure from EMC  (HP is EMC's largest customer).

>>2. SSA disks will correctly interact with all HP diagnostic tools
>>including the CE utilities, as well as HP predictive support.
>>
>
>Yes, but will HP run them for a customer using SSA disks?
Yes they will, I have seen it done.

>these utilities (except for predictive) password protected?
Yes they are, but we all know how easy it is to get the password.

In addition (important) there are some SSA-related native diagnostic
programs that are available, and are free.
>>
>>5. If you need features, buy EMC.  If you need pricing or performance,
>>go with SSA.
>>
>
>I agree (except for the HP support issues :-)
>
>>6. Be careful about any representations with SSA performance on
>>MPE/iX.  On the HP-UX side, it is pretty easy for SSA to double or
>>triple performance of Nike, EMC, or Jamaica.  On the 3K, because how
>>the O/S just hates to go to disk, you *might* see little performance
>>gain.  I could go into this in detail, but the information is very
>>system-specific.  I don't want to say that your IBM salesrep might be
>>misleading you on performance expectations, but you need to talk to
>>somebody who really knows both SSA and how it works with MPE/iX.
>>There aren't a lot of them out there.
>>
>
>Thank you, thank you, thank you!!  I'm delighted that someone selling and
>supporting third-party disk products for MPE/iX officially recognize that
>MPE/iX itself does wonderful things with disk I/O and that it is very hard
>to improve.  While some enviroments may see performance gain, the best
>reasons to look at non-HP disk solutions are more for fault tolerance/high
>availability/footprint/ etc. than performance.
You're welcome.  I try not to sound to biased, and shoot straight.  We
have seen some prospects gain as little as 3% improvement on some
tests (but their 995 with over 150 disk drives did an aggregate 500
IOPS, which was only 3-4 IOPS per drive per second.  Obviously they
didn't have an I/O problem which could be solved by SSA).


>>9. HP CSY spent over a month exercising and testing a large SSA
>>configuration on a 9x9.  They tested both the IBM 7133, as well as our
>>own (non-IBM) SSA disk subsystem.  Both subsystems passed their tests,
>>after a few firmware changes were made to the SSA interface adapter.
>>
>
>I sure would like to hear more from HP about this.  I have several
>questions.  First, did they spend a month exercising and testing or did a
>month lapse during their exercising and testing period?
Several months lapsed before they did anything (sigh).
Yes, they beat the heck out of it.
We tested multiple initiators, and I can't remember, but I think they
had 80 or so disk drives.  All the tests you would expect including
hotswap, adding 32 disk drives to the same FWD controller, and
mirroring were tested thoroughly.  Applications ranged from image to
predictive to other diagnostics not known to the general user
community.


Just exactly how
>hard did they work the devices?  What kind of throughput did they
>experience?  And so on...

>Hopefully, no-one sees my continuing intransigence regarding recommending
>SSA as concerns about the product and technology.  My concerns are strictly
>support and, as co-chair of the Interex High Availability Forum, I will
>continue to,
>1) Recommend solutions to HP users that represent the "wisest" (IMHO)
>choice for their environments and, 2) Encourage HP to develop relationships
>with other third-party disk providers that would allow market competitive
>pressures to force prices down and features and functionality up.  As a
>performance person, I'll continue to throw cold water on those third-party
>disk providers (not David) who communicate to HP users that their solution
>will result in fantastic performance gains in an MPE/iX environment.  It's
>just not possible, in most situations.
>
>
>Bill Lancaster
Agreed.  There is additional information on the saga of HP "support",
as well as HP's future plans, but the appropriate forum is not a
public media such as a newsgroup.

In addition, another technology that should be investigated is
external cache that can be used with HP's existing drives.  We've been
able to bring latency down from 7-10 ms down to 150-200 microseconds
which can make a much bigger difference in performance for certain I/O
problems.  I hope to be able to publish before/after results within 30
days.  We have already run "before" numbers through the wonderful Lund
Performance Audit.

David