Reg no.of cpu's,RAM

[boomer_2]

Hi guys,
We have setup of 2 rx4640 nodes in cluster with hpux 1i v2 ...8 gb ram each and 2 cpus each..
When we do EOD for banks for 1 bank our cpu utilizn goes to 70%..and now our PM told us that by next year we would be having 50 banks..Then should i upgrade my no. of cpus to 4 each and similarly ram from 8 to 32 gb each....will it be enough for our producn server ???

[Peter Godron]

Hi,
there is just not enough info here.
70% for how long ?
How long is your bank EOD run ?
Can you fit them sequentially into a batch run overnight or do they have to run at the saem time ?

For example:
If your load is 10 minutes for 1 run, then you could run 50 runs in 8.5 hours.

Please also read:
http://forums1.itrc.hp.com/service/forums/helptips.do?#33 on how to reward any useful answers given to your questions.

[Steven E. Protter]

Shalom boomer,

Most likely this is due to the way the application was written. If it is not efficient or does not permit itself to be moved from cpu to cpu, by being cpu bound, it will caused unbalanced cpu use.

SEP

[boomer_2]

Hi peter,
70% for how long ?
30 min...
How long is your bank EOD run ?
30 min...
Can you fit them sequentially into a batch run overnight or do they have to run at the same time ?

Will have to ask my Apllication team...

[A. Clay Stephenson]

You haven't even begun to ask the right questions yet. You really need to build a test lab so that you can accurately simulate the load. Moreover, eventhough you are running HP-UX, it might not be as good an answer as, for example, a cluster of Linux servers --- or HP-UX might be perfectly suited. How do the data arrive at your site? What kind of turnaround time is required?
What are your actual process bottlenecks at this point? Disk, Network, CPU, Memory, ... ?

And here is a very important question (and one that I would ask if I were one of your customers): What happens if your operation is down? How much redundancy is built into your system? Is there a failover site?

[boomer_2]

Hi Clay,
We do have a DR site of the same configuration...i.e. 8 gb ram,2 TB storage,2 cpu's....with 10 Mbps line from our primary to DR site..
I think ur right reg disk bottleneck,cpu utilizn etc..i really need to test it out for my application....
Thanx for ur answers guys...

[Hein van den Heuvel]

>> will it be enough for our producn server ???

Maybe yes maybe no. "it depends".

It is _likely_ that you can run more EOD's for more banks at the same time, but maybe you can only close one at a time with the current software in which case more cpu's will not help at all.

A more important question might be what the penalty will be if you get it wrong?

Will you get chance to fix it, or will the company fail to deliver on serivce expectations and loose business?

CPU is NOT at 100%, so there may be an other bottleneck imminent when more load is applied.

If this is at all important for your company, then the problem is well beyond the scope of a public forum and warrants a full fledged scaling and size excersize, possibly just on paper but more likely requiring a benchmark of sort.
The question will take anywhere from 2 weeks to 2 months to generate a worthwhile answer.
Please advice your management of the growth options with the current hardware and recommend a full study, possibly with help of a consultant (like myself :^) to manage the future growth.

Regards,
Hein van den Heuvel
HvdH Performance Consulting

[boomer_2]

i have upgraded memory and everythings fine..........