Hi there

I run SETI, Rosetta and WCG in my Intel Mac. The projects have equal resource shares. For some days now, Rosetta stopped downloading WU and now the same happened with WCG. Only SETI continues to work normally.
When I try to manually update the project, I get these messages below.

|rosetta@home|Sending scheduler request to https://boinc.bakerlab.org/rosetta_cgi/cgi
|rosetta@home|Reason: Requested by user
|rosetta@home|(not requesting new work or reporting completed tasks)
|rosetta@home|Scheduler request succeeded

Any ideas?

---

Try reducing your machine to two projects only - It's a common "BOINC" problem that it doesn't seem to cope well with more than 2 projects per machine.

Or, you can force it to get some project data by suspending all other projects and updating your Rosetta project at that time, but that will have to be done every time it runs out of work for a particular project.

I don't really understand the reason for this, but I've seen it before.

--
Mats

---

Try reducing your machine to two projects only - It's a common "BOINC" problem that it doesn't seem to cope well with more than 2 projects per machine.

Or, you can force it to get some project data by suspending all other projects and updating your Rosetta project at that time, but that will have to be done every time it runs out of work for a particular project.

I don't really understand the reason for this, but I've seen it before.

--
Mats

It is designed to work that way to prevent you getting so much work at once that you spill over the deadline. Remember in most projects (Rosetta and CPDN excepted) late work risks getting no credit!

Which project is "banned" (called negative long term debt) will rotate around your projects if you leave it to sort itself out. It even copes well (over a period of months rather than days) with uneven project shares, banning the large-share projects for less of the time than the small-share projects. To do this it has to be quite complex, and because the software takes a long term view, in the short term view of the human mind it often appears counter intuitive (= daft). Leave it alone and it will come right.

If you keep suspending to make it do what you think it should, it will never come right.

If you really don't like the software doing this, then it is possible to run as many projects as you like withou tmeeting this problem, so long as you keep the "connect every N days" value small enough.

To avoid getting into No Work Fetch mode, you need progressively smaller cache sizes as the number of projects grows.

If the recommended cache is C, the number of projects is N, and the shortes deadline of any project is D, then my rule of thumb is

C < D / (N+2)

John Keck recommends a different formula, which is

C < 0.4 * D / N

Try both and see which works best for you!

Of course, once you get so many projects that C days is shorter than a single task, then No Work Fetch becomes inevitable. But it is certainly possible to have more thAN 2.

On a box with 3 projects and a 1 week deadline, my rule of thumb suggests 1.4 days, and John's rule 0.93333 days. In my tests, with a setting of 1.4 days it occasionally goes into NWF for a few hours, but not for days on end (which I reckon is a success). With a setting of 0.93333 it never goes into NWF (which some others would call a success).

Hope that helps
River~~

---