From: Lukas Beeler
CPW, in my opinion, are an almost useless marketing number.
I find CPW to be quite significant in that it offers historical perspective - going back to about 1989 when the first "B" models were built. It seems that these days a lot of benchmarks are redefined after 3-5 years, so you lose historical continuity.
The problem is that CPW are only for machines with everything
maxed out (memory, disk, etc.).
That may be what IBM says. But if you compare CPW to CIW (compute intensive workload) values, both published in system handbooks, you see a strong correlation between the two. They both seem to correlate strongly to CPU performance.
Unfortunately, IBM doesn't publish benchmarks of the 1 core
machines running IBM i, making comparison difficult.
Ironically, the link you provided DOES list the 1 core machine, and shows both the AIX rPerf rating and the IBM i CPW rating. Also notice the strong correlation between those two ratings - across multi-core servers too.
Currently, a decent laptop ships with a 7.2kRPM SATA drive, 2
cores at 2-3 Ghz and 4GB of RAM.
Yes, well 64-bit Windows Vista is extraordinarily CPU & memory intensive. It's misleading to compare any kind of laptop to any kind of server.
the single core makes the machine rather laggy when
multiple people are accessing it ...
The advantage of the single-core machine is in the cost of software licensing. For example, MS SQL Server is priced at $25K per "processor", and most folks run it on a quad-core server.
IBM on the other hand offers a single-core model 520 for $6-$7K. You get a fully functional enterprise class database, but it's constrained by just one active core. Nevertheless, it's an exceptional value from a functional & cost of ownership perspective.
Now i'm rambling about IBM's pricing again, which isn't really
what i wanted to. So i'll leave it at that.
You seem to be suffering from the myopia that affects most people - only able to see the immediate out-of-pocket cost of a CPU, a Gig of RAM, or a hard drive. It takes a little bit longer to understand the total cost of ownership of consolidated workloads on an IBM i server, but it's worth it.
i5/OS eats memory for breakfast ...
If you look more closely you'll probably see that IBM i is NOT eating memory. The native virtual machine is remarkably efficient. I'm running over 100 different web-based database maintenance/transaction processing applications under the native virtual machine on a 1 GIG server, with exceptional performance.
A JVM instance, or a Java-based application on the other hand ...
Nathan.
midrange.com
