Also, out of curiosity, can't BOINC users configure BOINC to crunch other projects once one runs out? What would the disadvantages be for the crunchers if that happens? We obviously prefer to keep you here :D I just want to know why it is a problem if we shortly run out of WU's.

Yes, crunchers can crunch for other projects. You can set their weighting to be very low, or zero, and only get the odd task, or only get tasks if the main project runs dry.

You will find that many crunchers purchase GPU's to facilitate their choicest project. For a start, if we wanted to crunch for some other projects we would buy ATI cards, and while NVidia's work on some of these projects their relative performance to ATI cards is poor. MilkyWay@home is an example of such a project. Obviously Boinc credits are important for many, so if you use a GPU that is not suited to a project you will have reduced credits. Some projects might not be stable, or there could be compatibility issues running some mixes of work.

Other projects also run dry, and come to an end (WCG and POEM no longer have GPU WU's). You might need to configure to have 3 or 4. So for example, you might have Einstein, MW, and Albert as backup projects. This can be a lot of hassle especially if many projects frequently run dry.

If your project reputation deteriorates you become perceived as an intermittent project and lose many, if not most, of your crunchers. Crunchers come and go, but even those that would hang around cant stay at a project with no work! When they leave they might not come back for a long time.

What you could do is have another Long queue, and that way always have a backup project here. Send the work out from that project/queue as low priority; only if the other work is low or unavailable.

IMO you should be trying to model some more fundamental bio-techniques, and thus better establish your working model.

---

FAQ's

HOW TO:
- Opt out of Beta Tests
- Ask for Help

Also, out of curiosity, can't BOINC users configure BOINC to crunch other projects once one runs out?

Yes, we can. This is maybe my personal obsession, but I don't have any spare GPU project. I've built my rig especially for GPUGrid. Sure I could crunch with them for some other GPU project(s) as well, but then I could have built my computers using cheaper AMD GPUs. So their best use is here, at GPUGrid.

What would the disadvantages be for the crunchers if that happens?

My computers provide the heating of our apartment. So in the winter we could feel cold if there's no work on them :) (Don't worry, we have natural gas heating also, and we use it when it's really cold outside.)

We obviously prefer to keep you here :D I just want to know why it is a problem if we shortly run out of WU's.

The annoying part of it is that the RAC is dropping very quickly when it happens. But no real harm is done.

But I'd like to turn this question the way around.
You're using a 1.3PFlop supercomputer, which is granted to you free of charge. Its building cost is around 455.000€, and its monthly running cost is around 17.000€ (probably more, as there are older GPUs also). I would go mad, if I couldn't feed this monster with work. Don't you feel the same way?

Wow, guys. It is seriously not a big deal, if a project runs out of work. It's your decision on whether you want your computer to be able to be "kept busy" in that scenario. If you do, you can add other projects. And if you only want to crunch those other projects when your main project runs out of work, then set those other projects to a Resource Share of 0.

Want to know something interesting? This outage triggered a bugfix in BOINC!

In my 3-GPU computer, I had previously set up my 2 main GPUs to exclude even the backup projects. But when GPUGrid went dry, I realized that was the wrong setting for me, because I wanted to keep those GPUs busy.

So, in my quest to set things exactly like I wanted, I found a bug. Check out the following email exchange. :) The bugfix won't land until the next version of BOINC is released, but it's good to have been found and fixed.

Anyway, again, it all boils down to: If you want to keep your resources busy even when a project goes idle, then add more projects, and set the settings appropriately! For reference, I'm attached to 29 projects, and have the settings set such that the 2 main GPUs focus on GPUGrid.

Keep up the good work, GPUGrid, and if you run out of work sometimes, so be it.

==================================================
On 19-Mar-2014 3:37 PM, Jacob Klein wrote:

David,

I'm seeing a behavior, and I'm having trouble deciding if it's unsupported functionality, or if it's a bug. Can you please answer and help?

- On the BOINCstats BAM! account manager website, for one of the projects (SETI Beta), I have set a host-specific Resource share (to 0 in my case)
- This is a host-specific setting, separate from the "global project" resource share value.
- I tell BOINC to communicate with BAM!, and I see my client Resource Share get updated (RS changes to 0)
- I click Update on the project, and because the project does not have a "Resource Share change" queued up, my client still correctly says 0.
- I then go to the project website, and change the Resource Share, to 2 for instance
- I click Update on the project, and this time, because the project DOES have a "Resource Share change" queued up, my client now says 2.

Shouldn't it have respected my host-specific account manager value of 0? Or is the concept of a host-specific value, unsupported functionality?
For reference, the client sees that the Project is indeed managed by the Account Manager (I cannot click "Remove")

Please let me know,
Thanks,
Jacob

PS: For anyone who cares to hear the detailed reason of why I'm even playing with this, it goes like this:

- My setup was: 2 beefy GPUs set to work on GPUGrid, with other projects excluded, alongside 1 tiny GPU that had GPUGrid excluded. Exclusions are all done using <exclude_gpu> config in cc_config.xml
- That third GPU was set to only work Albert/Einstein/SETI/SETIBETA, but because I didn't want CPU tasks from them, and wanted to minimize RPCs, those 4 projects were set to Resource Share (RS) of 1.
- GPUGrid recently ran out of work, and my beefy GPUs were idle, due to my settings.
- So I changed the GPU Exclusions to allow the 2 beefy GPUs to work on the 4 projects, but because I wanted to prefer GPUGrid, I wanted to change the 4 projects RSs to 0.
- But I have a separate host with a tiny GPU that actually works those 4 projects and cannot do GPUGrid, and to minimize RPCs, I wanted to change the 4 projects RSs to 1.
- So I decided to leave the project RSs set to 1, but do an Account Manager host-specific setting of 0, on my main 3-GPU-rig, for the 4 projects.
- But that host-specific setting of 0, is getting trumped, and changing to a 1. Bug, right? Sigh.
- If anyone has a better way to accomplish what I want, without doing venues, I'm listening.

==================================================
Date: Wed, 19 Mar 2014 16:33:32 -0700
From: d...@ssl.berkeley.edu
To: boinc_alph...@ssl.berkeley.edu
Subject: Re: [boinc_alpha] Account Manager Host-specific Resource Share vs. Project Resource Share

It's a bug; if an account manager specifies a resource share,
that's supposed to trump the one returned by projects.

I fixed this (not tested).

-- David

==================================================
From: j...@msn.com
To: d...@ssl.berkeley.edu; boinc_alph...@ssl.berkeley.edu
Subject: RE: [boinc_alpha] Account Manager Host-specific Resource Share vs. Project Resource Share
Date: Wed, 19 Mar 2014 20:06:54 -0400

Thanks for the quick response. I'll try to give it some testing, in the next BOINC alpha release.
Host-specific settings are a handy feature for an Account Manager to have!

==================================================

Ok, thanks for the info :) So mostly dedicated equipment, RAC and messy BOINC configuration.
Yes of course we want to use it to it's maximum! And I am actually impressed at how much we do considering the amount of scientists in the lab. There is always though the "problem" that after simulating we need to analyze the data and publish :P And in parallel we need to send out new projects. It's tricky to time everything so well and sometimes the timing doesn't work out, like now.

I'm sickened by the number of "volunteers" who think this project should serve them instead of the other way around. "My RAC is falling, the sky is falling, I broke a nail, it's all your fault, now do the impossible and get me some tasks or I'm a whine like a baby, wah, wah, wah, wah, hurry up you're not doing the impossible fast enough".

I don't think I want to be part of a group that acts that way. I feel like I'm in with a bunch of 15 year old Twitter twits and bubble-gummers. I don't want anything to do with people like that. Thanks for all the advice and jokes but you're not the kind of people I want to be associated with.

Detached.

---

BOINC <<--- credit whores, pedants, alien hunters

I'm sickened by the number of "volunteers" who think this project should serve them instead of the other way around. "My RAC is falling, the sky is falling, I broke a nail, it's all your fault, now do the impossible and get me some tasks or I'm a whine like a baby, wah, wah, wah, wah, hurry up you're not doing the impossible fast enough".

I don't think I want to be part of a group that acts that way. I feel like I'm in with a bunch of 15 year old Twitter twits and bubble-gummers. I don't want anything to do with people like that. Thanks for all the advice and jokes but you're not the kind of people I want to be associated with.

Detached.

I think it's more the times, almost everything we do not is instant, a cup of coffee on every corner, a food place too, internet with ALL of it's info is at our fingertips 24/7, need a new part for something it can be delivered tomorrow! And then when there is a problem we want it fixed NOW too, just like everything else in our lives. We have forgotten about THE PEOPLE behind the scenes making everything work, unless of course we our selves are involved in those problems or fixes, or whatever. Want to file a FOI request, no problem here's the form. You say you want to wait for the response, GIVE ME A BREAK, it can take MONTHS or YEARS for it to be even considered let alone filled! People have forgotten what it takes to make this an 'everything at our fingertips World', the data is available and we want it NOW!!!

ps I HOPE you didn't REALLY leave, but if you did good luck in your next endeavor!

Dagorath:

Not everyone acts as you describe. I think it's a bit silly to leave a project due to the way other people act. But it's your call.
I'm here for the science.

- Jacob Klein

That felt so good I decided to get rid of BOINC entirely including browser bookmarks to anything to do with BOINC. I'm sure the feeling is mutual but tell someone who cares.

Expunge my records, please. Don't want anybody knowing I was here, damages my reputation as a thinker.

---

BOINC <<--- credit whores, pedants, alien hunters

That felt so good I decided to get rid of BOINC entirely including browser bookmarks to anything to do with BOINC. I'm sure the feeling is mutual but tell someone who cares.

Expunge my records, please. Don't want anybody knowing I was here, damages my reputation as a thinker.

Sounds like a bad day, I HOPE some soda/beer/whiskey/time/whatever will let you reconsider your decisions. But if not it was nice talking to you and I wish you the best!

.............Anyway, could those of us still here have some more WU's please?

Thanks!

We will receive more GPUGrid work, when more work is available.

In the meantime, I recommend attaching to multiple GPU projects, so that you can still keep your GPUs busy until then. You can even set their Resource Shares to 0, which will only get work when your main project(s) are empty.

............ And all flowing again.

Many thanks!!

In 2-3 days we will have lots of new simulations. Noelia and Nate are setting them up right now.

Let Noelia take her time! We don't want bad, hastily made WUs from Noelia! :)

---

Thanks -- let Noelia and Nate know that we appreciate any opportunity to volunteer for them!

I'm sickened by the number of "volunteers" who think this project should serve them instead of the other way around. "My RAC is falling, the sky is falling, I broke a nail, it's all your fault, now do the impossible and get me some tasks or I'm a whine like a baby, wah, wah, wah, wah, hurry up you're not doing the impossible fast enough".

I don't think I want to be part of a group that acts that way. I feel like I'm in with a bunch of 15 year old Twitter twits and bubble-gummers. I don't want anything to do with people like that. Thanks for all the advice and jokes but you're not the kind of people I want to be associated with.

Detached.

Wow..... that must be a world record distance for spitting the dummy out!!!

GPUGRID must be one of the most reliable BOINC projects and is a much more worthwhile project than a lot of them. A lot of people have invested a lot of money in expensive GPU cards to run this project and "Credit" and "RAC" is almost addicitive,

Alright!! It looks like we're actually going to run the server dry of GPU tasks again - Congratulations everyone!

I hope you've got your backup projects in place. I know I do!

I've got the following projects ready to accept GPU work:
- World Community Grid (which currently only has CPU apps)
- POEM@Home (very rarely has GPU tasks available)

... and I've got the following 4 projects set at 0 resource share, for exactly this situation, so my GPUs stay busy while I don't have any work available for my main projects:
- Einstein@Home
- SETI
- Albert@Home
- SETI Beta

So, remember, instead of complaining about lack of GPUGrid work, treat it as an accomplishment (we killed it - hurray!), and also keep those GPUs busy (to further serve humanity!)

Regards,
Jacob Klein

If the queue is empty it is also a good moment to let the rigs cool down, remove any dust everywhere, replace noisy fans, check the wiring, etc.

---

Greetings from TJ

TJ:
That's not a bad idea. Do a little spring cleaning on the inside of the case, then put them to work again. Thanks for the tip :)

And I thought I would keep you fed for a while with my BARNA WU's ;(