LHC@home

Hi!

I might be beating a dead horse, but could you please raise the daily quota? Especially with these relatively short WU's, my dual-cpu PC can chew through more than it is allowed to download pr. day. Maybe you could set the limit to X WU's per CPU - not per host?

> Hi!
>
> I might be beating a dead horse, but could you please raise the daily quota?
> Especially with these relatively short WU's, my dual-cpu PC can chew through
> more than it is allowed to download pr. day. Maybe you could set the limit to
> X WU's per CPU - not per host?

Per-CPU quota is not supported by BOINC so it's unlikely that we would do it. Increasing quota is problematic in the sense that some older machines can then download too much work at once (and possibly never finish it). But we'll think about it.

Can you give an estimate on how many workunits your dual-CPU machine can process in a day?

---

Markku Degerholm
LHC@home admin

> Per-CPU quota is not supported by BOINC so it's unlikely that we would do it.
> Increasing quota is problematic in the sense that some older machines can then
> download too much work at once (and possibly never finish it). But we'll think
> about it.
>
> Can you give an estimate on how many workunits your dual-CPU machine can
> process in a day?
>
Dear Markku

Afaik, he has Athlon MP 2800+ cpu's....
They should be comparable to my 244 Opterons... And my machine does about 80 WU's per day.

I do not have that problem. I have enough work left... :-)

---

This is mostly an issue to people that are single project oriented.

And I am most assuridly not saying that is wrong, just that it does "bend" the BOINC concept a little.

We still do not have a real good mechanism to allow people to customize their participation well (yet).

Like:
1) LHC@Home - run @ 50% when work is available

2) CPDN - download and maintain minimum 1 WU per CPU as insurance, normal share 30% of time left after LHC@Home has its share

3) Einstein@Home 35% after LHC@Home has its share

4) SETI@Home 50% after 1-3 have their share

5) Predictor@Home 50% after 1-3 have their share

or something like that. I probably won't see that for a week or two ...

---

> This is mostly an issue to people that are single project oriented.
>
> And I am most assuridly not saying that is wrong, just that it does
> "bend" the BOINC concept a little.

Dear Paul.

I agree. And I haven't complained about it. I have at least 1 alternative project. And one thing: If my host runs out of work... my power bill relaxes a bit... might want to keep that in mind as well.

Cheers,

Sysfried

---

Can you give an estimate on how many workunits your dual-CPU machine can process in a day?
==========

Any one of my P4 HT CPU's can due a minimum of 50 per day running in HT Mode even if they are all the longer ones ... Of course this is for the shorter ones & not the boince type WU's... I've noticed I'm starting to run into a problem with keeping enough WU's on each PC also ...

I thought the Daily Quota was at 100 or even more & then all of a sudden I started getting the Exceeded Daily Quota Messages on all my PC's ...

Maybe Cern will also add a line with your 'daily-quota' as PAH does.
(I would appreciate that.)

So you know from which number you talk, ...
I report in average more than 100 WU's/day with my HT-box (stats from boinclogX)
It depends also how many 's16_' and 's18_' you got to crunch.....

greetz littleBouncer
calculated: 0.6 hour/WU on a HT = 80 WU's/day....

Markku:

"Increasing quota is problematic in the sense that some older machines can then download too much work at once (and possibly never finish it)"

Isn't deadlines, benchmarks and estimated cpu time used to prevent this?

---

> Markku:
>
> "Increasing quota is problematic in the sense that some older machines can
> then download too much work at once (and possibly never finish it)"
>
> Isn't deadlines, benchmarks and estimated cpu time used to prevent this?

They should but I'm not sure if they actually work that way. Can anybody confirm..? In addition, estimates are only good as long as a BOINC CC is doing work 100% of time. If the BOINC client is shut down, there is no much that can be done.

---

Markku Degerholm
LHC@home admin

Well it's confusing as heck to just what the daily quota is supposed to be anyways ... I only have 7 PC's & 4 of them got 100 WU's today ... 1 got 70 & 2 got 50 before they received the Exceeded Daily Quota Message ... ???

> Can you give an estimate on how many workunits your dual-CPU machine can process in a day?

I use a dual Athlon MP 2800+ with 1GB Ram since Monday for LHC (100%) and it computed 280 results within that time. => 280/3 days = 93 WUs per day

But every day this PC reached the quota (~16 times per day), so there is the possibility to compute more WUs.

> Isn't deadlines, benchmarks and estimated cpu time used to prevent this?
I think that's right: On the first day my client downloaded about 90 WUs after benchmarking.
Other machines download less than that: I use a 600MHz machine too and that only fetches 10 WUs.

I think that the quota can be doubled without problems.

Another solution: doesn't the scheduler know how much WUs were sent and received? So when there are zero left, it can send more.

---

<a href="http://lhcathome.cern.ch/team_display.php?teamid=478">
<img src="http://boinc.mundayweb.com/lhc/stats.php?userID=294"></a>

> Per-CPU quota is not supported by BOINC so it's unlikely that we would do it.
> Increasing quota is problematic in the sense that some older machines can then
> download too much work at once (and possibly never finish it). But we'll think
> about it.
>
> Can you give an estimate on how many workunits your dual-CPU machine can
> process in a day?
I dont have the exact number, but each WU usually takes 35 minutes, so somewhere between 80 and 90 per day

dear Markku

It looks like, it's my turn to ask for higher daily quotas

The superfast machines above, thats the high end world.

Depending very of the actual processing stuff, me to, hitting the limit.
This not only with the faster hosts,
also with host of class pentium 2.66, several 3.0 GHz and the amds 2400/2800 are also offering the message.
They still do have some work, but can't "refuel"

The fastest host running, shoes this messages first time short bevor 12h00
now it 13:10 and this host has right now only 4 or 5 WUs (from LHC)left.

Now this host has to "wait" until midnight (nearly 12hours from now) until he got his "refuel".
It's rather cold outside, so this single cpu based Intel Host is running with 3.905 GHz right now.
The average time for a "normal" WU for this host is about 44 Minutes. It takes 2 at time (HT mode) so the returning is 8 WUs done in 3 Hours,
(the theoretical troughput is at least 64 LHC WUs per 24 Hours, not counting the very short ones).

Even faster single cpu based hosts are hitting with out problems the limit,
several slower hosts are also hitting the limit.

May I subscribe to the wishlist of higher limit, please?

At least until the mountain of WUs still to be processed, is done.

regard

ric

The ideal situation would be for the scheduler to keep statistics on how many WU a given machine has been able to process in a day and then supply use that amount of WUs and a bit more to allow for fluctuations as the quota. The problem is that it will take far too long for the scheduler to handle those request.

cheers,

---

Chrulle
Research Assistant &amp; Ex-LHC@home developer
Niels Bohr Institute

> The ideal situation would be for the scheduler to keep statistics on how many
> WU a given machine has been able to process in a day and then supply use that
> amount of WUs and a bit more to allow for fluctuations as the quota. The
> problem is that it will take far too long for the scheduler to handle those
> request.
>
> cheers,
===========

Another problem with trying to do that is the great variance between the length of time it takes to run the different s types (s6 through s18) ... The scheduler would have to know how long each one is going to take to for each individual WU I would think in order to send the proper amount of WU's for each CPU ...

And since each s type can take a different amount of time to process theres no way for the scheduler to send the correct amount of WU's ...
A definite Quagmire to say the least ... hehe

> That’s strange. According to Bruce Allen, on Einstein “The 'Daily Result
> Quota' is normally 8 workunits (per CPU, with a 4 CPU maximmum)”

Interesting. Either a) I'm wrong b) we are using too old version of the scheduler or c)instein project has done custom modifications.

Guess I need to find out which:)

---

Markku Degerholm
LHC@home admin

> > Per-CPU quota is not supported by BOINC so it's unlikely that we would do
> it.
> >
>
> That’s strange. According to Bruce Allen, on Einstein “The 'Daily Result
> Quota' is normally 8 workunits (per CPU, with a 4 CPU maximmum)”
>
> http://einstein.phys.uwm.edu/faq.php#daily_quota

THe daily quota is a project set number.

On Einstein@Home: Maximum daily WU quota 8/day
Predictor@Home: Quota 65
Seti@Home: Maximum daily WU quota 100/day

This number also can fluctuate up and down, with client side problems causing the system to lower the number because of the failures to complete work. Conversly, successes will raise the number to the cap, which should be the number listed herein ...

The number is now part of the Computer Summary screen and should be set by the later generation scheduleers.

---

> They should but I'm not sure if they actually work that way. Can anybody
> confirm..? In addition, estimates are only good as long as a BOINC CC is doing
> work 100% of time. If the BOINC client is shut down, there is no much that can
> be done.

I suspected as much with the estimates. In the end though I guess the uptime percentage messurements for each host are ment to take this into consideration, at least to further make sure clients don't get more work than they can handle in time.

---

> I suspected as much with the estimates. In the end though I guess the uptime
> percentage messurements for each host are ment to take this into
> consideration, at least to further make sure clients don't get more work than
> they can handle in time.

THe real problem is that it is something of a random walk...

We have an estimate of the runtime of an unstable simulation so that, the real processing time is one second to actual run completion time.

We have the computer on time, resource share, fate, ill-winds, and other daemons.

:)

When I was still working one of my tirades was "the estimate is not the schedule". In software development whenever an estimate is given, that is the OUTSIDE limit on the development time ... sigh ...

ANd I see from the front page we are blowing through the work too ... drat! :)

---

HM, my faster boxes are also out of work:

Host ID 26239

Host ID 31646

And, these boxes aren't allowed to bring their full power ! With HT, they have each 4 processors, but BOINC may only use 2 !

---

Supporting BOINC, a great concept !