The discussion regarding processing in WinBUGS has been very informative
for me, and hopefully others. Coincident with suggestions of some in this
recent thread, myself and a colleague conducted a benchmark assessment of 4
different machines (this has been done before, but a bit of time has passed
I guess since it was last reported to the group). Attached is a PDF
describing this test relative to the specifications of the machines.
The Dell Precision M50 was, one year ago, the best machine we had at our
Center (costing in excess of US$4000). The Dell Optiplex GX400 was a very
good machine at the time it was purchased ~2 years ago. In the course of
the last year, Dell has begun providing better clockspeed for the buck. We
found in our benchmark assessment that the newer machines performed an
order of magnitude faster than machines only 1 or 2 yo. What took us 43.5
hours, now can take us less than 4.5 hours. Thus, rather than investigate
the development of a 'LinBUGS' approach, we've decided to devote our
resources to hardware upgrades (A recent configuration for a nicely
appointed Dell machine with a 3.2 GHz processor costs US$4000; this is
substantially less than what it would cost to develop requisite software
and hardware and the incumbent maintenance). We've also decided, based
upon Simon Frosts' work, is to consider a distributed-processing situation
whereby we harness idle machines. Our Center is largely empty at night, on
the weekends, and on holidays, and so we may have upwards of 100 computers
available at any one time to process data. Parsing out chains to different
machines and then assembling them post hoc seems like a worthy endeavor to
pursue. I still like the idea of cluster computing, but it doesn't seem
immediately necessary for WinBUGS processing.
(See attached file: PC Specs.pdf)
Wayne E. Thogmartin, PhD
Statistician (Biology)
USGS Upper Midwest Environmental Sciences Center
2630 Fanta Reed Road
La Crosse, WI 54603
608.781.6309
[log in to unmask]
www.umesc.usgs.gov/staff/bios/wet0.html
-------------------------------------------------------------------
This list is for discussion of modelling issues and the BUGS software.
For help with crashes and error messages, first mail [log in to unmask]
To mail the BUGS list, mail to [log in to unmask]
Before mailing, please check the archive at www.jiscmail.ac.uk/lists/bugs.html
Please do not mail attachments to the list.
To leave the BUGS list, send LEAVE BUGS to [log in to unmask]
If this fails, mail [log in to unmask], NOT the whole list
|