Fewer, faster processors

[Mark Tunnell]

Does anyone have any information regarding the comparative performance of an N4000 with four 440 Mhz processors compated to an RP5470 with two 750 Mhz processors? The application is an Oracle database, reading only. In general, are two faster processors roughly equivalent to 4 slower (half the speed) processors?

Thanks,

Mark

[Sanjay_6]

Hi Mark,

I would not think that they are equivalent.

This link may help,

http://www1.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000062689038

The itrc doc id is CSY9301150015.

Hope this helps.

Regds

[A. Clay Stephenson]

You are making the classic mistake of evaluating the processors --- and leaving out the other variables. In the majority of cases, database servers are i/o bound rather than CPU bound so that there are cases where systems that differ greatly in CPU throughput may only vary slightly in actual performance because the CPU is such a small component of the picture.

In very general terms, I would expect the CPU performance of the two to be similar. Because of the synchronization overhead of each additional processor, I tend to choose fewer, faster processors.

[Geoff Wild · Accepted solution]

What I have found (cause I went through this earlier this year) - we moved from a 8 cpu 440 MHZ Nclass to 4 x 750 RP7410.

The CPU load used to be 60 to 80 % on the N, and on the rp7410 - 80 to 95%!

Why? My understanding is, 1 cpu is basically dedicated to OS - so what really happened was we went from 7 to 3!

Think of it in terms of a water pipeline - you had 7 pushing out say 440 litres per sec (max) each, and now 3 at 750 litres per sec (max).

What happens when you want to push 100 litres per second? not much - both pipeline can handle it...

WHat if you want 2000 litres per second? Well the 3 way pipeline is almost jammed!

In the end - we added a 5th cpu...

Another analogy - 7 ok workers and 3 faster ones - yet, as the workload increases - the faster ones can't keep up...

Rgds...Geoff