[Bioclusters] Apple/Genentech BLAST

A.J. Rossini bioclusters@bioinformatics.org
24 May 2002 12:44:28 -0700


>>>>> "ivo" == Ivo Grosse <grosse@cshl.org> writes:

    >> pay more per-node for a queuing environment) but I'd appreciate any
    >> other suggestions on free cycle stealing queuing environments.

    ivo> Could anybody who has experience with Condor and Mosix compare the two? 
    ivo>  Of course, the concepts are different, but comparable in the sense 
    ivo> that both allow you to make use of idle computers in the network.  I 
    ivo> would like to learn which of the two

    ivo> - is less intrusive,

MOSIX.

    ivo> - is easier to install,

Tie (depends on if you prefer kernel mucking with user/system-level
config mucking)

    ivo> - is easier to maintain,

MOSIX (except for kernel issues; you might have to roll-your-own patch)

    ivo> - is more robust,

Condor (Checkpointing is a plus)

    ivo> - is more efficient, i.e., has less overhead,

Not clear -- MOSIX generates a fair amount of network traffic, and
occassionally (maybe "rarely" is better) 

    ivo> - cause less network traffic.

See above.  I'm not sure that Condor does network evaluations of CPU
usage/loads as often, but MOSIX works on a local-level, i.e. figures
out load on nearest neighbors, for balancing.

    ivo> Are there other alternatives?

Condor is heterogeneous, for a limited definition of hetergeneous;
MOSIX currently is only stable (beta or non-alpha) on i386 Linux
platforms.

Condor does scheduling; MOSIX could be configured to schedule amount
of usage, for "most" of the cluster (but has no way of limiting job
activity by itself -- a bit of code (scripting) would be required to
be written to set up policy for used, and possibly backgrounding jobs
based on users, etc.

I really think that they are semi-orthogonal (CPU load balancing could
result in weird things if both are active and aggresively balancing
the same set of processes on a homogeneous cluster, but I've got no
data (or experiences) to prove this).

best,
-tony

-- 
A.J. Rossini				Rsrch. Asst. Prof. of Biostatistics
U. of Washington Biostatistics		rossini@u.washington.edu	
FHCRC/SCHARP/HIV Vaccine Trials Net	rossini@scharp.org
-------------- http://software.biostat.washington.edu/ ----------------
FHCRC: M-W: 206-667-7025 (fax=4812)|Voicemail is pretty sketchy/use Email
UW:   Th: 206-543-1044 (fax=3286)|Change last 4 digits of phone to FAX
(my friday location is usually completely unpredictable.)