>> Can anyone tell me the approximate percentage gain in going from HP-UX v1 to v3?
No.
Unless you want a vague -10% to +20%
It depends too much on your application.
Even adding something like 'oracle based' or 'web serving' is not going to give enough common ground to work from.
Here is a hint though: For starters look for yourself at how much SYSTEM time your application uses versus USERmode.
Kernel/OS changes are rather unlikely to improve usermode time, so if you application is doing 90% usermode, then there is very little hope of significant improvement through OS changes/tuning.
exceptions: - Kernel change could change translation buffer usage and context switing and such making caches more effective, reducing user mode time).
- The DBC was improved which would be system time, but coudl indirectly improve user time (less waiting / spinning / scheduling).
Potential performance gains are a nice side effect of going V3, but should never (rarely) be the deciding factor.
Support for new hardware, new software tools, new certification boundaries, new patches, better support status, those are the real reasons to move forward.
For example, you might for to v3 in preparation of going to more cost effective Itanium hardware. Or maybe a new Oracle release needs a more recent HPUX.
You are right though to keep an eye on performance.
Be sure to have a reproducable, reportable, long term performance overview established under v1 to compare with for v3.
Anything will do, even a crude vmstat script, but seriously look at 'sar' and 'measureware' and commercial tools for comprehensive reporting if you do not already have this in place..
Hope this helps some,
Hein van den Heuvel (at gmail dot com)
HvdH Performance Consulting
