LHC@home

This may sound a little bit weird as some of you guys are dying to have your daily quotas increased, but I'm simply downloading too many workunits. Let me explain:

My cache is set to 1 day, and completion time is estimated to be 1h20m. I'm downloading 18 WUs at a time, so that makes exactly 1 day of work. But I have 4 other projects running, and L@H's resource share is 10%, so I should really be downloading 2-3 WUs at most at a time. It used to be like this in October and November before the looooong silence.

So is this recent influx of WUs a rush to send out as many WUs as possible to test something so that resource shares and client uptimes aren't taken into account? Ot is this a bug in the scheduler?

You probably know this already but all you have to do is just set your Preferences to 0.1 or 0.2 days to get less ...

Yes, I know that, but that will affect my other projects, too. I found my resource shares and cache settings after a lot of experimenting. They are perfect for me for the time being in that I can finish my WUs before deadlines AND my results are mostly included in the quorums (I don't feel I'm contributing to science if they are not!)

I don't want to change this, and besides other projects take my resource shares into account and don't fill my cache with 1 day of WUs each. I just want to learn why L@H is not behaving like other projects do.

More than likely your times have been changed for completion with the LHC WU's, this could give you more WU's, lowering the resource share for LHC is about all you can do then to get less from LHC...This won't affect the other Projects then...

HM, the resource-share had and has a lot of points to know / to recognize:

1) It depends on the actual client you are running.
2) There have been a lot of changes to the design of resource-share
3) With older clients, it was by design, that, if a project had no work for some time, after delivering fresh work, the client crunched more from this than from other projects.

Actualy, the development-clients are changed, so that they give resource-share, Queue deep and deadlines of projects a much better balance. But, it can take several days until this balance is found ...

---

Supporting BOINC, a great concept !

@ Yeti
> But, it can take several days until this balance is found ...
>
>
>
For that you need that all projects, you are attached to, are running smoothly!!!
What is not often the case:-(

greetz littleBouncer

> So is this recent influx of WUs a rush to send out as many WUs as possible to
> test something so that resource shares and client uptimes aren't taken into
> account? Ot is this a bug in the scheduler?
>


Well, According to Project Admin Markku Degerholm in this thread: http://lhcathome.cern.ch/forum_thread.php?id=1338

The LHC Client is using an older Scheduler and they are in the process of Alpha Testing a newer replacement. Maybe this has to do with your problem???

> So is this recent influx of WUs a rush to send out as many WUs as possible to
> test something so that resource shares and client uptimes aren't taken into
> account? Ot is this a bug in the scheduler?

If it's a bug, it's a bug in the core client. The scheduler only has a maximum limit (quota) for results to download. It's up to the core client ask for proper amount of work. Which version of core client you are using?

---

Markku Degerholm
LHC@home admin

4.25

> > So is this recent influx of WUs a rush to send out as many WUs as
> possible to
> > test something so that resource shares and client uptimes aren't taken
> into
> > account? Ot is this a bug in the scheduler?
>
> If it's a bug, it's a bug in the core client. The scheduler only has a maximum
> limit (quota) for results to download. It's up to the core client ask for
> proper amount of work. Which version of core client you are using?
>
With everything after 4.20, the CC always asks for the queue size and leaves it up to the server to do the math. (This would not be my first choice of how to do this). We may have to switch back to having the server just believe the client because the correct amount of work requested should also take into account the LT debt.

---

BOINC WIKI

> > > So is this recent influx of WUs a rush to send out as many WUs as
> > possible to
> > > test something so that resource shares and client uptimes aren't
> taken
> > into
> > > account? Ot is this a bug in the scheduler?
> >
> > If it's a bug, it's a bug in the core client. The scheduler only has a
> maximum
> > limit (quota) for results to download. It's up to the core client ask
> for
> > proper amount of work. Which version of core client you are using?
> >
> With everything after 4.20, the CC always asks for the queue size and leaves
> it up to the server to do the math. (This would not be my first choice of how
> to do this). We may have to switch back to having the server just believe the
> client because the correct amount of work requested should also take into
> account the LT debt.
>
Amen there. The scheduler has enough to do without second guessing the clients work requests. Of course the clients still need to make better work requests so that the server side computations are redundant. That's why I wanted requests to be made in CS or flops, it removes another calculation that can be done on the client but is currently being done on the server.

---

BOINC WIKI

BOINCing since 2002/12/8