LHC@home

I'm running 5 clients, and none of them are sending me any work. When I updated them all, I got:

7/22/2005 7:31:04 PM|climateprediction.net|Sending scheduler request to http://climateapps2.oucs.ox.ac.uk/cpdnboinc_cgi/cgi
7/22/2005 7:31:04 PM|climateprediction.net|Requesting 0 seconds of work, returning 0 results
7/22/2005 7:31:04 PM||request_reschedule_cpus: project op
7/22/2005 7:31:04 PM||request_reschedule_cpus: project op
7/22/2005 7:31:05 PM||request_reschedule_cpus: project op
7/22/2005 7:31:07 PM|climateprediction.net|Scheduler request to http://climateapps2.oucs.ox.ac.uk/cpdnboinc_cgi/cgi succeeded
7/22/2005 7:31:08 PM|Einstein@Home|Sending scheduler request to http://einstein.phys.uwm.edu/EinsteinAtHome_cgi/cgi
7/22/2005 7:31:08 PM|Einstein@Home|Requesting 0 seconds of work, returning 0 results
7/22/2005 7:31:09 PM|Einstein@Home|Scheduler request to http://einstein.phys.uwm.edu/EinsteinAtHome_cgi/cgi succeeded
7/22/2005 7:31:10 PM|LHC@home|Sending scheduler request to http://lhcathome-sched1.cern.ch/scheduler/cgi
7/22/2005 7:31:10 PM|LHC@home|Requesting 0 seconds of work, returning 0 results
7/22/2005 7:31:11 PM|LHC@home|Scheduler request to http://lhcathome-sched1.cern.ch/scheduler/cgi succeeded
7/22/2005 7:31:12 PM|ProteinPredictorAtHome|Sending scheduler request to http://predictor.scripps.edu/predictor_cgi/cgi
7/22/2005 7:31:12 PM|ProteinPredictorAtHome|Requesting 0 seconds of work, returning 0 results
7/22/2005 7:31:13 PM|ProteinPredictorAtHome|Scheduler request to http://predictor.scripps.edu/predictor_cgi/cgi succeeded
7/22/2005 7:31:14 PM|SETI@home|Sending scheduler request to http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi
7/22/2005 7:31:14 PM|SETI@home|Requesting 0 seconds of work, returning 0 results
7/22/2005 7:31:15 PM|SETI@home|Scheduler request to http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi succeeded


What am I doing wrong?

---

Probably nothing. It depends on how much work there is on your computer, how that work is distribued between the processes and how much CPU time has been used by each project recently.

There are 3 causes for a request of 0 seconds of work:

1) A host that is already overloaded with work.

2) A project that has enough work and a host that is not desperate for work.

3) A project that has had more than its share of CPU time recently.

---

BOINC WIKI

Like Sarah, I have zero second requests ... but only from the LHC process.

After several "Update"s, I tried "Reset"ting the project ... which kind-of worked - reloaded all project software and 4 WU's. After all 4 WU's completed, back to zero requests!!!

Did the "reset" thing again with remakably similar results.

So, LHC is doing nothing at present but SETI and ClimatePredictor are chugging away quite happily.

Anybody got any ideas????

---

Chas - Orme's Tun, Mercia, Albion.

<blockquote>Like Sarah, I have zero second requests ... but only from the LHC process.

After several "Update"s, I tried "Reset"ting the project ... which kind-of worked - reloaded all project software and 4 WU's. After all 4 WU's completed, back to zero requests!!!

Did the "reset" thing again with remakably similar results.

So, LHC is doing nothing at present but SETI and ClimatePredictor are chugging away quite happily.

Anybody got any ideas????</blockquote>
This is bad advice. It just delays the problem.

If you let it run it should get to a more or less steady state. It will take an estimated 2* queue size * number of projects before it settles. If you have a slow enough computer with a large enough queue it will never settle, but will run some projects for awhile then change to other projects and ect.

---

BOINC WIKI

BOINCing since 2002/12/8

Another problem that will keep the program from ever getting into a steady state are the projects that have very short deadlines.

And another is just a large number of projects attached to with even a very short queue (every computer has a limit that will drive it into EDF and No Work Fetch).

---

BOINC WIKI

Thanks, John,,

I suspect this is the most likely cause ...
<blockquote>....
3) A project that has had more than its share of CPU time recently.
</blockquote> ...
so my next question is "How does this happen?" - or more to the point, how do I stop it happening? ... what parameters in my preferences can I tweak?

---

Chas - Orme's Tun, Mercia, Albion.

<blockquote>Thanks, John,,

I suspect this is the most likely cause ...
<blockquote>....
3) A project that has had more than its share of CPU time recently.
</blockquote> ...
so my next question is "How does this happen?" - or more to the point, how do I stop it happening? ... what parameters in my preferences can I tweak?</blockquote>
Reducing your queue size is the best thing you can do. That reduces the chance that a project will get more than it's fair share of CPU time. My general advice is take the longest you expect a local outage (not project) to last and divide that by the number of projects you are attached to.

---

BOINC WIKI

BOINCing since 2002/12/8

<blockquote>Thanks, John,,

I suspect this is the most likely cause ...
<blockquote>....
3) A project that has had more than its share of CPU time recently.
</blockquote> ...
so my next question is "How does this happen?" - or more to the point, how do I stop it happening? ... what parameters in my preferences can I tweak?</blockquote>
And for some project/computer combinations, it cannot be avoided at all. For example I have one slow computer that is attached to BOINC Alpha and SZTAKI. SZTAKI has a deadline of 4 days, and this computer takes about 3.5 days to complete a result for SZTAKI, so every time a SZTAKI WU is downloaded, the computer will enter EDF and No Work Fetch - just guaranteed. However, if it continued with Round Robin, the work would not be completed on time.

---

BOINC WIKI

<blockquote>... Reducing your queue size is the best thing you can do. That reduces the chance that a project will get more than it's fair share of CPU time. My general advice is take the longest you expect a local outage (not project) to last and divide that by the number of projects you are attached to.</blockquote>

Thanks, John K, (this could get confusing - too many Johns?)

To clarify (for my simple mind) - I originally set up BOINC for SAH (when the server went "missing" every weekend) to buffer the workload on my system) and set up prefs: Connect to network about every 4 days (determines size of work cache; maximum 10 days) reckoning that would give me enough WUs to keep going.

Since then, I have joined ClimatePrediction and LHC - both with the 4 day "work cache".

As I undersatnd it, what I should do now is to cut back on the "work cache" size (1 day?) so that any "missing server" is covered by the WUs for "other" projects ... right or wrong?

---

Chas - Orme's Tun, Mercia, Albion.

<blockquote>
As I undersatnd it, what I should do now is to cut back on the "work cache" size (1 day?) so that any "missing server" is covered by the WUs for "other" projects ... right or wrong?</blockquote>
Correct

---

BOINC WIKI

BOINCing since 2002/12/8

<blockquote>... Correct</blockquote>

Thanks, again, John.

Going totally-off-topic ... looked at your profile and thought I had a magic mirror - "That's me, twenty years ago ... or is it Kev?" - if you haven't looked at my profile, Kev is my doppleganger pal who works at CERN, so I guess that means your are a trippleganger?

---

Chas - Orme's Tun, Mercia, Albion.