[Beowulf] Re: Why Do Clusters Suck?
Robert G. Brown
rgb at phy.duke.edu
Wed Mar 23 11:56:48 EST 2005
On Wed, 23 Mar 2005, Joe Landman wrote:
> > might require too much revision for BBS to remain unbroken, though, so
> > we may yet need to create "son-o-BBS".
> I have been working up a roadmap for it, and the first B needs to be
> dropped or exchanged in favor of other things. Somehow, I thought
> Cluster Benchmark System might elicit some cease and desist letters from
> lawyers, so still thinking on this ...
Feel free to co-opt plain old benchd or xmlbenchd. Or XBS -- XML-based
This might create a desireable level of abstraction between the "daemon"
-- which is simply a tool that can run the shell that runs the
benchmark(s) according to policy, extract the results and put them into
some sort of database (to be determined, actually), and on demand return
those results via the daemon interface -- and the shell itself that runs
the benchmarks, which might be and XBS compliant microbenchmark binary
(after we finish a draft of the unified XML and I finish hacking
benchmaster to comply with it) or an XBS-script-wrapped non-compliant
microbenchmark binary (e.g. stream or lmbench or netperf of netpipe) or
an XBS-script-wrapped macro benchmark, a.k.a. "any old application,
ideally but not limited to your own" (e.g. BLAST or linpack or a
partitioned lattice problem written in MPI or PVM) done with wall clock
BTW, at some point this discussion may becoming boring or offensive as a
"waste of time" to list members. I'm hoping not yet, as this is VERY
USEFUL to me as I seek to figure out what should be in the toolset and
how it should be split up into (as Don points out) RELATIVELY simple and
disjoint tools with distinct functionality and an ABI between them, but
if so just yell and we can try to set up an offline group or list (which
I'm sure we'll do soo anyway).
Robert G. Brown http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567 Fax: 919-660-2525 email:rgb at phy.duke.edu
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
More information about the Beowulf