RS/6000 Multiuser Performance

[Ruth Milner]

In article <19136 at rpp386.cactus.org> jfh at rpp386.cactus.org (John F Haugh II) writes:
>In article <78003 at bu.edu.bu.edu> jdh at bu-pub.bu.edu (Jason Heirtzler) writes:
>>Last week I posted a similar question, but noone responded with more
>>than 65 simultaneous users (not xterminal users, but "generic timesharing"
>>users, if there is such a thing) so let me ask again: Does anyone have
>>an RS-6000 with more than 50 simultaneous users?  If you do, your input
>>would be greatly appreciated!
>
>I regularly use a S/6000 with upwards of 120 users on it.  The machine
>is used by IBM support staff and contractors to access a database of
>APAR and "How To" question information.

You really need to be more specific here, i.e. which *model* of RS/6000 are
you using?

Someone asked about using a 320 to serve a bunch of Xstations. I have no
experience with Xstations, but I do know that the interactive response time
on a 320 is abysmal with more than a couple of users on it, especially if
one of them is running anything remotely CPU- or IO-intensive.

The 530 is not bad, the 540 is quite good. Never used a 550 or any of the
other high-end systems. It depends so much on what your users do, though,
that the question is very difficult to respond to generally.

BTW, when you're configuring any RS/6000, bear in mind that the kernel is
huge compared to other UNIXes (unices?). We got our 320 with 16MB, normally
just fine for our application, and when we found performance to be very
disappointing IBM loaned us a 16MB board to try. Result: the application
ran on average 1.5x faster, in some cases twice as fast as before. The
reason was quite simply because the AIX kernel used half the original 16MB.
-- 
Ruth Milner
Systems Manager                     NRAO/VLA                    Socorro NM
                            rmilner at zia.aoc.nrao.edu