[GRLUG] large systems performance

[Bob Kline]

On Nov 16, 2007 1:16 AM, Michael Mol <mikemol at gmail.com> wrote:

> On Nov 16, 2007 12:57 AM, Tim Schmidt <timschmidt at gmail.com> wrote:
> > On Nov 16, 2007 12:45 AM, Bob Kline <bob.kline at gmail.com> wrote:
> > >  I'd think swapping could be pretty fast in an
> > > out of memory, but the instant it involves
> > > a hard drive it's over.
> >
> > ?  Swapping always involves block storage of some type.  How much it
> > hurts again, depends on memory access patterns of your load, as well
> > as what percentage of the working set is swapped, and so on.
>
> Heh.  It could start hurting even more.  Apparently they're getting
> close to implementing swap over NFS.  It's intended for diskless
> cluster environments, but it still stands to complicate performance
> issues.  Sub-optimal performance?  Do you have a bad network cable, or
> are your jobs eating too much memory?
>
> --
> :wq
>
>
Defining efficiency on a system is
complicated anyway.  In order to
create the illusion of many users
each able to run a job on a machine
at the same time there has to be
overhead introduced.  In many cases
the most efficient efficient mode is
one job running.  The trick is to be
somewhere between one job running
and all overhead.

Something like a Linux version of
the Laffer curve.

     -Bob
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://shinobu.grlug.org/pipermail/grlug/attachments/20071116/52759fdd/attachment.htm