These are machines that don't meet the memory requirements for a workunit.

My question: Does this mean there is absolutely no work available that the machine qualifies to run, or does it mean just not qualified to run what happened to be sent to it?

---

Currently, the project has two types of WUs, some require 256MB memory, some require 512MB. Your machines are hidden, so I can't tell. Do you have 256MB of memory?

I for one am unclear exactly how the system handles these memory limitations. Are you getting any messages in BOINC Manager as you attempt to download work? Do you know if a WU was actually downloaded before reporting the message?

I believe BOINC is working on some enhancements to the scheduling to help better accomodate such situations and deliver work that's appropriate to the client.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

Currently, the project has two types of WUs, some require 256MB memory, some require 512MB. Your machines are hidden, so I can't tell. Do you have 256MB of memory?

I for one am unclear exactly how the system handles these memory limitations. Are you getting any messages in BOINC Manager as you attempt to download work? Do you know if a WU was actually downloaded before reporting the message?

I believe BOINC is working on some enhancements to the scheduling to help better accomodate such situations and deliver work that's appropriate to the client.

Cant remember where, but someone proposed that if u set a second project of choice up, e.g. Tanpaku or Simap and give it e.g. 0.01 factor of time share, Rosetta will be treated as primary. If there is no work that fits, the machine starts pulling work from alternate projects, until Rosetta has something on offer again. WCG has a few life science projects coming with very small needs, so have a sniff there too. In fact WCG restarted HPF2 which is running on the Rosetta principle, 1 WU, many seeds, no quorum. Drs.Bonneau well known to Dr.Baker is the architects of that work. It's expected to take a long time for that project to complete. HPF1 of same took 27,000 CPU years to complete.

---

Coelum Non Animum Mutant, Qui Trans Mare Currunt

Currently, the project has two types of WUs, some require 256MB memory, some require 512MB. Your machines are hidden, so I can't tell. Do you have 256MB of memory?

I for one am unclear exactly how the system handles these memory limitations. Are you getting any messages in BOINC Manager as you attempt to download work? Do you know if a WU was actually downloaded before reporting the message?

I believe BOINC is working on some enhancements to the scheduling to help better accomodate such situations and deliver work that's appropriate to the client.

One is 256 the other is 384. The machine with 384 is running on Knoppix Live CD which is probably making it worse. The machine needs a hardrive. I'm not sure if it actually downloads the WU before the error is given.

---

I've got quite a few machines with <512MB RAM, although some of these use less than 80MB for everything other than Rosetta.

what's the official line on this - anything in the pipeline?

cheers
Danny

---

Sekerob Just to clarify, there's no magic to having a second project with a small resource share. BOINC has no concept of "primary project". And this isn't going to help the problem originally reported. Although it would certainly be a way to assure the machines have SOMEthing to do (i.e. the other project).

Having a second project configured just gives BOINC another server to look for work on when it needs work to do. And it just follows its normal debt and resource share rules. And these rules basically say to get work for the project with the highest resource share (actually highest debt, which is tallied based on resource share)... if you CAN... and if there isn't any, it will try the another project.

So it ends up working out as though there is a priority to the projects. But its really just the normal tracking of time per project as compared to resource share. As the project with high resource share runs longer and longer, the other project is accumulating a time debt. If the other project ever gets a single WU, it will likely take longer then the debt to crunch it, and so then the secondary project won't try to get work again for some time.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

Serkob,
WCG-HPF I & II are quite well know here, since the updated HPFII client was improved here first. I believe they use the version we helped improve during the first 1/2 year or so of R@H.

I think you mean
... Rosetta will be, in effect, treated as primary ...
Truxoft for an earlier boinc client did add project priority to Rosetta@home (source on his site) but no one I know of has added some of his features to larter clients. [which is a shame]

There are some added client logging features
https://boinc.bakerlab.org/forum_thread.php?id=2550 I posted about that may point out the decisions the scheduler is making and why you are not getting any more tasks. Maybe the <work_fetch_debug> option ?

---

Team mauisun.org