I would make the cautious proposal that, if you're looking at
consolidating from Intel or Sparc to zSeries, that, if CPU (rather than
I/O or memory bandwidth) is really your bottleneck, the project is
probably not a good candidate for consolidation.

In short, and in my experience, grain of salt, all that....if what you
need is more computing cycles, your most cost-effective option is Intel,
assuming that it's a parallelizable workload.  If it's not then you're
probably looking at something like a pSeries, for extremely high
per-processor performance.  If disk or memory bandwidth is your
bottleneck zSeries may be a very good choice.  If you're not hitting any
constraints and can't easily tell which wall you're going to hit first,
there's probably not an effective near-term case for consolidation in
any event.

If infrastructure management is your primary bottleneck--and this
requires thinking outside the box a bit, since it's not something you
can directly measure on the machines--then zSeries may be an excellent
candidate for you.

Adam

Reply via email to