Trying to get my BOINC host to download WUs


log in

Advanced search

Questions and Answers : Unix/Linux : Trying to get my BOINC host to download WUs

Author Message
Team AUSTRALIA (AlexD)
Send message
Joined: 1 Jun 99
Posts: 54
Credit: 66,602
RAC: 0
Australia
Message 2975 - Posted: 1 Jul 2004, 3:32:17 UTC
Last modified: 1 Jul 2004, 7:05:32 UTC

As I understand, unlike S@H classic, BOINC is able to "batch download" a configured number of days worth of WUs and queue them for processing upon request. Have been successful in getting my BOINC host to download WUs for itself, but for the life of me I can't work how to link my S@H Classic (soon to become BOINC) "powerhorses" to it!

A bit of background:

Have a "server farm" of some 12 Solaris-based SunFire 280R 1015Mhz dual-processor machines running some transaction processing software which has primative clustering abilities. I say primative because it can't load-share across all the Sun boxes, rather it fires transactions at the first box until it says "I'm busy", then targets the second, and so on. In this way, it's "transactions per minute" rate grows as demand increases, at peak times all 12 machines can be working hard, but these come in infrequent bursts, there's many more times when up to 8 or 9 machines are totally idle (other than housekeeping).

Ah ha! Perfect fodder for S@HC (S@H Classic), the admin boys wrote a little cron task which starts or stops S@HC (depending on if there's demand from the transaction process or not (think it's based on CPU load).

Anyway, these S&HC "powerhorses" cannot, will not and shall not ever have access to the Internet, not even via proxy, firewall or carrier pigeon (management paranoia at it's best, and not totally unjustified in this case). So we have an old Windows 98SE Pentium 3-600Mhz PC which does have Internet access as our S@HC host, setup with SetiQueue to do the WUs queuing (and it processes WUs in quite reasonable time too).

However, now that BOINC can perform this batch downloading of WUs and with its new communications protocols, we can't use SQ anymore or need to, instead stick to using BOINC exclusively. However, after downloading the Solaris version of BOINC (boinc_3.18_sparc-sun-solaris2.7.gz), it only seems to contain the executable, no readme text with details on command-line parameters or what proxy port to use etc. Any help would be appreciated, even if it's a link to where the documentation is.

TIA

Darren
Volunteer tester
Avatar
Send message
Joined: 2 Jul 99
Posts: 259
Credit: 277,051
RAC: 2
United States
Message 3211 - Posted: 1 Jul 2004, 16:04:36 UTC

"As I understand, unlike S@H classic, BOINC is able to "batch download" a configured number of days worth of WUs and queue them for processing upon request. Have been successful in getting my BOINC host to download WUs for itself, but for the life of me I can't work how to link my S@H Classic (soon to become BOINC) "powerhorses" to it!"

If you're trying to distribute a workunit from your windows system to one of the solaris systems, this can't be done. Unlike the queues in seti classic, a queue in boinc can only be used by the instance of boinc that created that queue.

There was some talk in beta of possibly implementing some type of farm queue in the future, but when (or even if) is unknown. As it is now, each wu is logged to a specific computer and must be completed and returned by that same computer. If any of the steps produce an identity other than the one specified at the time the queue is created, the results will be rejected upon return. No doubt you could ultimately manage to move a work unit and create the needed xml file to get it to run on a different computer, but the end result would be that when it came back to Berkeley analyzed by a solaris system (that has never even created a host id within your boinc account) when it was assigned to your known windows system, it would be tossed.



Team AUSTRALIA (AlexD)
Send message
Joined: 1 Jun 99
Posts: 54
Credit: 66,602
RAC: 0
Australia
Message 3362 - Posted: 2 Jul 2004, 4:39:17 UTC - in response to Message 3211.

> If you're trying to distribute a workunit from your windows system to one of
> the solaris systems, this can't be done. Unlike the queues in seti classic, a
> queue in boinc can only be used by the instance of boinc that created that
> queue.
>
> There was some talk in beta of possibly implementing some type of farm queue
> in the future, but when (or even if) is unknown. As it is now, each wu is
> logged to a specific computer and must be completed and returned by that same
> computer. If any of the steps produce an identity other than the one
> specified at the time the queue is created, the results will be rejected upon
> return. No doubt you could ultimately manage to move a work unit and create
> the needed xml file to get it to run on a different computer, but the end
> result would be that when it came back to Berkeley analyzed by a solaris
> system (that has never even created a host id within your boinc account) when
> it was assigned to your known windows system, it would be tossed.

Bugger!!! No more impressive S@H processing for me for the short term it seems, as this means I'm reduced to the P3 and whatever my home PC can produce. I'm sure we're not alone in this predicament, so maybe we can use this forum as a way of discussing possiblities, and in this way, the Berkeley boys can follow along and add comments as appropriate.

Questions and Answers : Unix/Linux : Trying to get my BOINC host to download WUs

Copyright © 2014 University of California