LHC@home

At the moment we are doing tunescan workunits. This means we are trying to find the border region where the motion of the protons become chaotic. Especially the WU we loaded yesterday have very high amplitudes. This makes it very likely that particle go lost and therefore that WU with expected runtime of hours stop after a few minutes. Even though they do not run to the end of their turns the results are still very important to our analysis.

The shortrunning units does increase the bandwidth we need, but the load on the machine has not reached even 50 % and the network load seems very low compared to our bandwidth. I think we have simply allocated to few database connection, so the problem is that incoming requests are waiting for a connection. I'll put Markku on it as soon as i see him. The problems seem to be worst when we are loading jobs on the machine. We should get one machine more next week and we will look into moving the job submission on ton that machine.

Christian Søttrup
LHC@home Developer

still walking strong good job we will see what put to us ;-)
<a>[url=http://guidowaldenmeier.de]<a>

Christian

The problem I am having with these short work units is on my laptop which is a dialup connection. By the time a bunch of WU's have downloaded half of them are ready to upload. A few minutes later the rest have also finished. Not a problem on a cable connection however its getting to be a real hassle with dialup.

Hopefully I can pull down some long WU's soon for that host.

> The problem I am having with these short work units is on my laptop which is a
> dialup connection. By the time a bunch of WU's have downloaded half of them
> are ready to upload. A few minutes later the rest have also finished. Not a
> problem on a cable connection however its getting to be a real hassle with
> dialup.

I see your point but there is not much we can do about it right now, sorry. Maybe some random workunit selection thing could work - need to think about it.

Markku Degerholm
LHC@home Admin

> > The problem I am having with these short work units is on my laptop which
> is a
> > dialup connection. By the time a bunch of WU's have downloaded half of
> them
> > are ready to upload. A few minutes later the rest have also finished. Not
> a
> > problem on a cable connection however its getting to be a real hassle
> with
> > dialup.
>
> I see your point but there is not much we can do about it right now, sorry.
> Maybe some random workunit selection thing could work - need to think about
> it.
>
> Markku Degerholm
> LHC@home Admin

Yep, I guess well get to the other side eventually. Just something to think about for the future.

Cheers

Angs

We modified the feeder process such that it will give out work units in more random manner. We hope that this will prevent many subsequent short work units to be given out (it's still possible, but probability should be lower).



Markku Degerholm
LHC@home Admin

It looks like a lot of the Tunescan & now protwo Wu's are short ones, the 50 WU a day limit is killing me. 3 of the Computers I have will each run between 65 & 72 of the normal 40-50 min WU's a day. With a lot of the tunescan & protwo being short ones I can run out of work from the project in a half a day or less.
Sigh ;)

> Would it not be possible to either check under primary preference to
> see if a user answered yes to either confirm before connecting to
> internet or disconnect when done, or assign a flag to users
> who have so answered so that when they connect, they are prioritised to
> receive any 'larger' work units in the queue?

In principle, yes. But in practice we don't know in advance which work units may fall short and which don't - the number of rounds alone doesn't really tell the total run time because the simulated protons may collide to wall on the first round.

In optimal case, when estimated run time actually corresponds to actual run time, things should go fine because in the preferences users can tell how much they want work to be downloaded at a time.

Markku Degerholm
LHC@home Admin

> It looks like a lot of the Tunescan & now protwo Wu's are short ones, the
> 50 WU a day limit is killing me. 3 of the Computers I have will each run
> between 65 & 72 of the normal 40-50 min WU's a day. With a lot of the
> tunescan & protwo being short ones I can run out of work from the project
> in a half a day or less.
> Sigh ;)
>
Hmmm...that's what multiply projects are for. In fact attach to CPDN get one wu and resource share 80%lhc and 20%cpdn (but for me it's 55%cpdn and 45%lhc) and never run out of work again!

Happy crunching!

Paul.
[/url]

ps. Doh! Is this sig allowed?

This all assumes that there are actually some WUs available to download. :-)
Keep up the good work, Markku.

PeterV.

Hmmm...that's what multiply projects are for. In fact attach to CPDN get one wu and resource share 80%lhc and 20%cpdn (but for me it's 55%cpdn and 45%lhc) and never run out of work again!

Happy crunching! Paul.
=========

Hmmmmmmmmm Really, WoW thats Great News, I didn't even know there were other Projects, I'll have to try some of the them, Thanks for the Info Paul ... ;)

no short not long JUST ZERO ;-))
http://www.guidowaldenmeier.de

I hate to say it.. but mine still have work for another 24-48 hours... ;-)

Well I still have work on 4 PC's but it's really had to determine on how long they will last. Some run for 40-50 minutes & some only 4 or 5 seconds. For every one that seems to run the 40-50 minutes there are 3 or 4 that will only last 4 or 5 seconds ...