Condor Considerations?

[Mystwalker]

Has anyone ever tried to set up proth_sieve on a Condor cluster?

I'm in the beginning of doing so at my institute (20-30 machines, roughly 300G/day if running...).
So far, I managed to get some experience in it, but wasn't able to get it just right.

I'll continue trying in the next days and hopefully find out how it works. But maybe someone already knows how... :confused:

[Mystwalker]

Well, I think I finally figured it out. Just running 28 jobs, should yield up to 400G/day :D
Let's hope all goes well - I have only tested small ranges (1M) so far...

[royanee]

Awesome! By the way, I've been watching and wondering: How's 220000-225000 going? It's such a huge single range that I was wondering what kind of factors you are pulling out of it. :)

[Mystwalker]

Well, I've reserved that range when I still had direct control of those computers in the Condor Pool. But once that were set up, proth_sieve as a simple process wasn't allowed anymore, as it 'confused' Condor (using the whole computing power, Condor isn't able to use these PCs).
That's why it took a little longer than thought (having only ~3 systems available). When all goes right with Condor, I'll be ready in maybe a week or two at most (for now, I have to manually set up the ranges). Using only my own computers, it'll still take until End July, I assume.

Btw.: You can watch the progress of the range on MikeH's Gapa Analysis.

[royanee]

Thanks for the info! :) I knew about the gap analysis, but forgot that most people seem to send their factors as soon as they get them. For me the difficultly is remembering to submit the rest if I have already submitted some. I know keroberts had that situation with the p/k/n factor that could have saved a test that was assigned. But isn't that a rarity?

[Mystwalker]

It's very unlikely, right. Sometimes I have the feeling that all factors I'm finding are for n=10M+...:p

[Mystwalker]

Ok, seems like I found a way to use the Condor Pool - thus my range should be complete somewhen next week. :bouncy:

Still not a very elegant way, but I'm afraid Sieve Automation will be completed only after the exams (till end of July)...

[royanee]

Yay! Way to go! New stats leader? Sieve that range! :notworthy

[Nuri]

Originally posted by Mystwalker
I'm afraid Sieve Automation will be completed only after the exams (till end of July)...

That's enough room for a sieve of 17-18T, right? :elephant:

[Mystwalker]

Originally posted by Nuri
That's enough room for a sieve of 17-18T, right? :elephant:

Just to put it right:
When Sieve Automation (aka SobSieveServer) is finished, this won't take away resources from SoB Sieving - it will be a bit easier for me to control the work (though I just wrote some nice programs + script so I only have to do 2 commands for a new creation and insertion of jobs and 1 more to get all found factors :|party|: ) and maybe will make sieving for other people more comfortable as well.

But I'll soon evaluate putting in some Riesel or PSP sieving jobs, as they haven't sieved this deep as we did at SoB...
Defending #3 in the stats make sure I'll still put certain resources into SoB. :D