Forwarded message:
> From: Steve Guthrie <[EMAIL PROTECTED]>
>
> Sorry to confuse you.  I'm running under VM, of course.  As far as the size
> of the machine, why is that relevant?  I have an MP3000 H70, which is still
> small until you compare to Intel or Mac.  Then its big at about 115 MIPS.
> Unless someone gives me a bigger system, that's what I got.  I'm trying to
> optimize the system, running S/390, and Z/OS, as stated.  JIT is on,
> with -Xmx512m -Xms256m passed to the JIT -enabled VM.
>
> Now, any ideas or are we just looking for fuzzy little nits;-)?

   a)   An MP3000-H70 is, IIRC, a box like the P/390 in that OS|2
        is running to handle much of the I/O.  Yes, it's bigger
        than a PC and it's CP's are NOT those single-board s/390
        CMOS CPUs;  But it still won't win many prizes.

   b)   The -H70 has 2 CPs and is actually capable of being LPAR'd.

        I do not recall whether is has the "good" FP capability (IEEE
        rather than BHP).  I would imagine that you've LPAR'd this
        box into a z/OS (OS/390) partition and a z/VM partition-  and
        that you have Linux within a VM instance.

        Some issues:

   1)   Paging.  Is your Linux instance actually performing paging?

                There's a good chance that VM has to page in an object
                just so that Linux can page it right back out.

   2)   Paging.  Is z/VM doing any paging for the Linux instance?

                I/O buffer objects (along with computational pages)
                may be getting dragged back in.

        Mind you, the above will not really consume all that much CPU.
        Java runtimes don't tend to be very efficient at managing
        compute resources-  and if they're multithreading, you may find
        incredible load averages w/o seeing the CPU pegged to the wall
        (I see this often on AIX boxes-  especially if more than one
        JRE version is running in parallel with others).

        You really need to keep track of VM's own performance monitors.
        Not being a VM/CMS maven, I'm more that a little innocent when
        it comes to performance tools-  and Linux's own performance
        tools (vmstat, sar, etc) can't be trusted within a VM-ified
        instance (though an LPAR image will give you a better cut)
        (and I don't think an "absolute share" is that absolute, is it?).

        Now, have I shot my mouth off or what?

--
 John R. Campbell           Speaker to Machines                 [EMAIL PROTECTED]
 - As a SysAdmin, yes, I CAN read your e-mail, but I DON'T get that bored!
   Disclaimer:  All opinions expressed above are those of John R. Campbell
                alone and are seriously unlikely to reflect the opinions of
                his employer(s) or lackeys thereof.  Anyone who says
                differently is itching for a fight!

Reply via email to