Re: Namd benchmark problem

From: Nicholas M Glykos (glykos_at_mbg.duth.gr)
Date: Mon Oct 26 2009 - 09:14:36 CDT

Hi Axel,

> > I should add as a side note that -at least to my experience- there is no
> > single optimal way of tuning the operating system, such that it performs
> > best for all possible problem sizes and protocols. If you tune the OS
> > using the ApoA1 system, you shouldn't be surprised if it turns-out that it
> > is not optimal for, say, a peptide simulation. For large centralised
> > clusters, of course, you do not have a choice (which usually is a relief ;-)
>
> if you have a software on your operating system, that is _that_
> sensitive to two different systems in the same code, then there
> is something wrong with the OS software that needs to be tweaked.

Ok. An example then. Take PCI bus and a network card sitting on it:
changing the PCI latency of the card will decrease the latency, but will
also reduce maximum throughput. For a small MD system, the messages
exchanged are short, the bottleneck is definitely latency, and you are
better-off zeroing it (as seen on lspci). But for, say, a very large
system with a small number of nodes, the bottleneck may well be
throughput, in which case you want the card to hold-on to the bus for
longer (higher value in lspci).

> if one operates a machine for multiple purposes, the best option
> is usually to first characterize the dominant usage and tune a
> machine for that and then make sure the machine is usable and
> the performance does not degrade too badly for other uses.

Completely agree on that.

Regards,
Nicholas

-- 
          Dr Nicholas M. Glykos, Department of Molecular Biology
     and Genetics, Democritus University of Thrace, University Campus,
  Dragana, 68100 Alexandroupolis, Greece, Tel/Fax (office) +302551030620,
    Ext.77620, Tel (lab) +302551030615, http://utopia.duth.gr/~glykos/

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:51:37 CST