[ale] Performance

[Jeff Hubbs]

OK, I read you - but regarding the "disk IO on all the small files," is
this mostly a read operation or a write operation?  Reason I ask, how
would it be to leave at least the reading coming from disks on the
individual Mosix nodes, from kernel trees set up via rsync ahead of
time? 

When you did this, how did you establish the shared cluster file space?
NFS?  Mosix File System (something I need to look into again - it's been
a while)?

- Jeff

On Fri, 2003-11-07 at 12:32, James P. Kinney III wrote:
> On Fri, 2003-11-07 at 12:14, Jeff Hubbs wrote:
> Mosix did not perform well for me doing kernel compiles. The network
> latency was too high since there is so much disk IO on all the small
> files. Mosix really shines at parallel distributed applications that are
> number crunchers.  For a compile monster, SMP is the processor
> arrangement to use as it is able to use disk IO better. It is almost a
> linear speed up per CPU.
> 
>  SCSI Ultra320 RAID drives at 15K RPM on a PCI-X bus make nice disk
> subsystems :) Some motherboards have a dual PCI-X bus so one can spread
> the the load around onto a mirrored raid 5 system. That gets twice the
> read speed (but only single speed write).
> 
> > Dunno what your situation is, but if you do this A LOT and you have
> > other people there who are in the same place also doing it A LOT, you
> > might want to set up a mega-mutha box as a compile server OR, go the
> > Mosix route and make a compile farm out of a lot of cheapies or a
> > handful of biggies.
> > 
> > - Jeff
> > 
> > _______________________________________________
> > Ale mailing list
> > Ale at ale.org
> > http://www.ale.org/mailman/listinfo/ale
-- 
Jeff Hubbs <hbbs at comcast.net>