I cannot get more than ~6 WUs in the Tasks tab at any given time. My rig completes much more than that in a 24 hr period, but not close to the limit imposed by R@H (which I believe is 100 WUs/day)

Running BOINC v7.4.42
R@H is set at 20% Resource Share in the Projects tab.

In the preferences on website:
Computer is connected to the Internet about every
(Leave blank or 0 if always connected.
BOINC will try to maintain at least this much work.)
This is blank.

Maintain enough work for an additional
This was increased twice and is presently set for 2 days

---

Edit: After re-reading your post, I think my former reply was going in the wrong direction. This is probably more to do with the resource share of Rosetta vs. whatever other projects your crunching for. If you have 2 projects, really, the resource share between them should be set to 50-50 and then adjusted upwards/downwards from there over time until you're happy with the results. BOINC Manager handles the resource share thing a little lopsidedly if one project is disproportionately higher than another.

2 days of work on a machine that is to run 20% R@h is about 9 hours of work if the machine is on 24hrs/day. So depending how long you are running the tasks (see workunit runtime preference in the R@h settings via the website), that may be roughly one task per CPU.

Also, the BOINC Manager needs time (a few days usually) to adjust to changes in resource shares. It may need to complete some work for other projects before it is ready to download a full 20% schedule of work. It may also be adjusting to the length of your runtime preference. See if the estimated time to complete a task that has not started yet is close to your runtime preference.

The 100 tasks per day per CPU limit is really only relevant if the machine is failing tasks. Each success reported further extends the day limit.

---

Rosetta Moderator: Mod.Sense

@Timo
Yes, running 2 projects. I increased the share for R@h about 4 weeks ago, from 15% to 20%. This may not seem like much, but I was curious if this would have any bearing on WU cache. At 20% share, my RAC for R@h is 83% of the other project. Until a week ago, it was 76%. The change is due to implemented mods by the other project. Also, the proximity of the RACs in contrast to the difference in resource share most likely is (I’m guessing) due to higher credits granted by R@h b/c the WUs require more from users' hardware for computation.

Additionally, I’ve gleaned from various DC fora that BOINC does not have pinpount accuracy with such adjustments.

@Mod.Sense
9hrs/WU was a good # for my previous rig. {ASIDE: I haven’t been able to run R@h on it for some months now because, I believe, R@h deprecated SSE, and that rig was powered by an Athlon XP 2700. This was the last straw after an extended period of procrastination - I built a new rig.} My new rig has a multi-core processor which crunches a WU anywhere from 2.75-4hrs/core. This averages to ~40 WUs/day with
Target CPU run time = 4 hrs. (Would R@h benefit more if I increased that setting?)
Hence, my desire to increase the WU cache.

See if the estimated time to complete a task that has not started yet is close to your runtime preference.

This has always been aligned with preference setting.

BTW, as I write, the Tasks tab shows 10 WUs, which includes crunching and non-crunching. So, I will monitor it over the next several days and go from there.

Thanks

---

A few questions to help understand your environment.

Is the machine running BOINC 24hrs per day?
How many processor cores does the machine have?
How much memory per core is available for BOINC to use?
What is the other project you are running? Does it often run out of tasks to send?
I guess I am curious as well, is BOINC Manager more accurately retaining a 2 day cache of work for the other project?

---

Rosetta Moderator: Mod.Sense

@Timo
Yes, running 2 projects. I increased the share for R@h about 4 weeks ago, from 15% to 20%...

The share number isn't a %age figure. Every project defaults to 100. The Project tab in Boinc shows this on my machine.

I have Rosetta set to 2900 (96.67%), with WCG at 100 (3.33%) and currently have 20 x 6hr tasks on this dual core machine (0 of WCG), holding a 2 day buffer. On my 8-core machine I have ~45 x 8hr tasks.

Don't be shy with the share number. Changing from 15 to 20 is insignificant.

---

A few questions to help understand your environment.

Is the machine running BOINC 24hrs per day? Yes
How many processor cores does the machine have? 4 physical, 8 logical (Intel Core i7-6700K)
How much memory per core is available for BOINC to use? From Tools > Computing Preferences > disk and memory usage tab
use at most 50% when computer is in use (= 8GB)
What is the other project you are running? Seti@home
Does it often run out of tasks to send? Never
I guess I am curious as well, is BOINC Manager more accurately retaining a 2 day cache of work for the other project? I can’t state with the same conviction as with R@h, but I’d have to say no.

@Sid

@Timo
Yes, running 2 projects. I increased the share for R@h about 4 weeks ago, from 15% to 20%...

The share number isn't a %age figure. Every project defaults to 100. The Project tab in Boinc shows this on my machine.

I know, and I specifically entered #s so that R@h's share would = 15% (or 20%). Presently, I have the other project’s Resource share = 100, with R@h’s = 25, which yields 80% and 20% respectively.

Don't be shy with the share number. Changing from 15 to 20 is insignificant.

I thought it might.
What I’m trying to accomplish is to avoid continually contacting the R@h server for WUs, by increasing the time interval. In short, to lessen the traffic on the project’s server. If this would have a negligible impact on the server, I will understand. But, I’ll remain curious as to why this setting doesn’t function as expected.

Would R@h benefit more if I increased the Target CPU run time from 4 hrs?
Also yesterday, the WUs in the Tasks tab dropped from 10 to 6. Presently, it sits at 8, with 5 "Running" and 3 "Ready to start".

---

Would R@h benefit more if I increased the Target CPU run time from 4 hrs?

Yes, that would be another way to reduce number of scheduler requests. Make any changes to the setting gradually (one notch per day), because any tasks you currently have will change as well, and it will take the BOINC Manager a few task completions to assimilate the new runtimes into it's scheduling.

---

Rosetta Moderator: Mod.Sense

@Sid

@Timo
Yes, running 2 projects. I increased the share for R@h about 4 weeks ago, from 15% to 20%...

The share number isn't a %age figure. Every project defaults to 100. The Project tab in Boinc shows this on my machine.

I know, and I specifically entered #s so that R@h's share would = 15% (or 20%). Presently, I have the other project’s Resource share = 100, with R@h’s = 25, which yields 80% and 20% respectively.

Don't be shy with the share number. Changing from 15 to 20 is insignificant.

I thought it might.
What I’m trying to accomplish is to avoid continually contacting the R@h server for WUs, by increasing the time interval. In short, to lessen the traffic on the project’s server. If this would have a negligible impact on the server, I will understand. But, I’ll remain curious as to why this setting doesn’t function as expected.

Would R@h benefit more if I increased the Target CPU run time from 4 hrs?
Also yesterday, the WUs in the Tasks tab dropped from 10 to 6. Presently, it sits at 8, with 5 "Running" and 3 "Ready to start".

With so few tasks I wouldn't worry about hits on Rosetta's server. It's trivial in the grand scheme of things. I do agree with increasing the runtime to 6 hours (which is the project's new default). It would seem to address both your concerns at the same time.

I'd also look to increase Rosetta's share as a matter of principle, especially so when the other project is Seti.

Edit: I'd also upgrade the Boinc Manager to the current 7.6.22 - I think some task scheduling functions have been improved since your version. Scheduling has been a longstanding issue (that is, it's been rubbish for years).

---

With so few tasks I wouldn't worry about hits on Rosetta's server. It's trivial in the grand scheme of things. I do agree with increasing the runtime to 6 hours (which is the project's new default). It would seem to address both your concerns at the same time.

I changed Target CPU run time to 6 hrs two days ago. The CPU time/WU remains unchanged. The tasks within BOINC Msnager, at one point, jumped to 12 (Running + Ready to start) but decreased to 9. Additionally, this is after I changed the website BOINC preference
Maintain enough work for an additional 2 days to 4 days.

I'd also look to increase Rosetta's share as a matter of principle, especially so when the other project is Seti.

In all the time I’ve run both projects concurrently, I’ve seldom seen S@h CPU WU running - GPU WUs yes, and these only ran when the computer was not in use.

Edit: I'd also upgrade the Boinc Manager to the current 7.6.22 - I think some task scheduling functions have been improved since your version. Scheduling has been a longstanding issue (that is, it's been rubbish for years).

From reading the fora and through personal experience, scheduling (resource sharing) remains a weak link within BOINC. However, according to BOINC
http://boinc.berkeley.edu/trac/wiki/VersionHistory
upgrading for the sake of this one feature would be fruitless. I reviewed the entire 7.x series - nearly 4 years of releases.

---

I just noticed on the website preferences that under "Primary (default) preferences" the
Target CPU run time= 6 hours
while under "Separate preferences for home" this is set to 4 hours. Would this have any bearing on the initial work unit cache matter?

Would this have any bearing on the initial work unit cache matter?

Yes, the preferences are specific to the venue assigned to a specific host. The venue of a host is displayed in the host details, as well as in the BOINC Messages as the BOINC Manager is started.

---

Rosetta Moderator: Mod.Sense

I'd also look to increase Rosetta's share as a matter of principle, especially so when the other project is Seti.

In all the time I’ve run both projects concurrently, I’ve seldom seen S@h CPU WU running - GPU WUs yes, and these only ran when the computer was not in use.

Edit: I'd also upgrade the Boinc Manager to the current 7.6.22 - I think some task scheduling functions have been improved since your version. Scheduling has been a longstanding issue (that is, it's been rubbish for years).

From reading the fora and through personal experience, scheduling (resource sharing) remains a weak link within BOINC. However, according to BOINC
http://boinc.berkeley.edu/trac/wiki/VersionHistory
upgrading for the sake of this one feature would be fruitless. I reviewed the entire 7.x series - nearly 4 years of releases.

The line I was thinking of is in 7.6.9 where it says:

• client: fix job scheduling bug that starves CPU instances

I thought it might be relevant to your subject, but the choice is yours. It won't hurt anything and isn't a particularly big download.

My comment on Seti is more to do with my opinion on that project's futility.

---

I see that out of the 17 most recent WUs completed, 14 were ~6 hours in accordance with my recent setting change. Although, the # of WUs in the Tasks list remains unchanged @6.

I just noticed on the website preferences that under "Primary (default) preferences" the
Target CPU run time = 6 hours
while under "Separate preferences for home" this is set to 4 hours. Would this have any bearing on the initial work unit cache matter?

Yes, the preferences are specific to the venue assigned to a specific host. The venue of a host is displayed in the host details, as well as in the BOINC Messages as the BOINC Manager is started.

The following line is from the Event Log:
"General prefs; no separate prefs for home; using your defaults"
So, the difference between the above two settings should be moot, right?

The line I was thinking of is in 7.6.9 where it says:

• client: fix job scheduling bug that starves CPU instances

I thought it might be relevant to your subject, but the choice is yours. It won't hurt anything and isn't a particularly big download.

Just installed v7.6.22, and it caused a problem with the other project that needs to be fixed.

"General prefs; no separate prefs for home; using your defaults"

This means this computer will use the setting for home, i.e. you have to change it there or, if you actually don't need separate preferences for home, you might just delete them and it will use your default preferences. Note: we're talking here about your R@H computing preferences, BOINC preferences is another thing and have nothing to do with the runtime of your Rosetta WUs.

The line I was thinking of is in 7.6.9 where it says:

• client: fix job scheduling bug that starves CPU instances

I thought it might be relevant to your subject, but the choice is yours. It won't hurt anything and isn't a particularly big download.

Just installed v7.6.22, and it caused a problem with the other project that needs to be fixed.

This fix has nothing with the issue you have, it's about CPU going idle.

---

.