[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Distributed networks?
SETI@Home style distributed computing works if the
data in a "work packet" is quite small compared to the
amount of processing required. Or I should say, if
"the time to transport a work packet is small compared
to the time to process a work packet"
A perfect case for SETI@Home would a factoring. You
send out a 50 digit integer and ask that it be factored
into a set of primes. Much work is required but only
50 bytes of data is sent.
TASS is on the "wrong" end of this. You would have to
send out megabytes of data (several full sized FITS images)
and the processing takes only about 60 seconds.
The ratio of data transmitting to data processing is
not good.
With TASS the best solution is to "take the processing
to the data" not the "data to the processor". This is
true even after the data are proceessed. It's still best
to keep the data in a central database and if it needs to be
searched or plotted then you do that at the database site
and only ship the result back.
Data transport will remain expensive for a long time and I
don't see the data rates moving upward much past what it takes
to ship a movie into someone's house which is about a megabit
per second for a compressed stream. (Please don't argue that
you have a 6 megabit cable connection. I ment _sustained_
speed averaged over hours.) CPUs WILL get faster
over time. So to ratio if "time to transport" vs.
"time to process" will only get even worse as we wait and
SETI@home style processing be become even less usful except
in the odd casses like factoring huge integers
--- "Robert J. Bradbury" <bradbury@aeiveos.com> wrote:
>
> Something that might work to our advantage...
>
> The folks behind SETI@Home have produced a package
> called BOINC [1] that allows one to allocate fractions
> of your compute resources.
>
> As I don't don't think much of the SETI@Home project I've
> never used my CPU resources for that. Instead I prefer to
> dedicate them to Folding@Home [2]. If it ever gets off the
> ground I intended to use the BOINC platform for Nano@Home [3].
>
> The nice thing about BOINC is that it lets you turn up/down
> the fraction of your CPU(s) dedicated to various projects.
> So one could imagine lots of different people joining the
> TASS project (offering up CPU resources instead of
> attempting to manage the telescope & camera nightmares [at
> least from where I sit...]). Offloading the data processing
> might allow you to be a little more relaxed about when a
> machine goes down the tubes.
>
> When you run into a problem of machines going down you
> could just send a note to the TASS community to ask them
> to up their BOINC priorities for TASS. If enough people
> joined one could perhaps even eliminate home data processing
> (lowering your electricity and A/C requirements presumably...).
> You might have to shift attention a bit to redundancy in
> internet communications (Cable + Satellite?) and a router
> that can deal with that (a Linux gateway machine probably
> shouldn't have problems).
>
> Robert
>
> 1. http://boinc.berkeley.edu
> 2. http://folding.stanford.edu
> Note that I don't think F@H uses BOINC yet but I'm hoping
> they move in that direction...
> 3. http://www.nanoathome.org/
>
>
>
=====
Chris Albertson
Home: 310-376-1029 chrisalbertson90278@yahoo.com
Cell: 310-990-7550
Office: 310-336-5189 Christopher.J.Albertson@aero.org
KG6OMK
_______________________________
Do you Yahoo!?
Win 1 of 4,000 free domain names from Yahoo! Enter now.
http://promotions.yahoo.com/goldrush