(no subject)

Eric Miller emiller at techskills.com
Thu Apr 11 16:52:04 PDT 2002


<snip>
>And not even for SETI will you get an Nx speedup on N nodes.  There is
>ALWAYS a serial fraction even for embarrassingly parallel applications,
>and the time required to send the jobs out to the nodes
<Snip>

I guess what I am fundamentally saying is, for a cluster to be "working its
magic" it a student's eyes consider two scenarios:

1- Running N iterations of a program, and seeing work^N being done.  It's
like, um... well yeah if I run SETI on 8 systems, then I will crunch 8 times
as many units, but I will NOT crunch 1 unit in 1/8th the time as percieved
on the front end node.

-OR-

2- Having an 8 node cluster running, say, a raytracer.  Then, having a solo
machine running the same application.  Actually seeing ONE instance render
an image (roughly) 8 times faster than a single system (esp. when all of the
systems were pulled out of the trash can!!), THAT is the magic that newbies
and students want to see.  That's the "cool" factor that bring annoying
#^$%s like me to this forum and post questions that are outside the arena of
analyzing proteins and DNA molecules on 256 node AthlonXP rackmounts with
Myrinet.  We are not experts, we have ALOT of questions, and all we want to
do is see Linux do something cool that we can show our
freinds/students/selves.

Robert, thank you for your positive and informative reply.




More information about the Beowulf mailing list