Whenever Climate Predictor start to run, it prevent ANY other projects from running. It always run as \"High priority\". It completely ignore any resources sharing preferences.

Is there a way to enforce the resources sharing AS I set it?

If left unchecked, Climate Predictor will run for months on end, without ever stoping to let any other project any chance to run.
I need to manualy suspend Climate Predictor if I want other projects to do any work.

---

Hi,

Sounds like a deadline problem. Could you take a look at the deadline showing on your Boinc manager against the model, and see if it looks OK? Sometimes they get reset to 1901 (only on versions of the Boinc manager prior to 5.8.16).

If the deadline is OK, but near, and you have many projects, then the Boinc client may have decided that the project is getting insufficient resource share to complete within the deadline. You have a 160 year model, and have completed 9 model years since the model downloaded on 31st May, which means the Boinc manager may think that the model will take 7 years of wall-clock-time to complete. If you prefer, you might want to process a shorter model instead - if you go into \'project preferences\', you can select HadSM3 rather than HadCM3. The HadSM3 models are significantly shorter, and may suit your PC better.
http://climateapps2.oucs.ox.ac.uk/cpdnboinc/result.php?resultid=6578810

There are also a lot of crashes listed against your AMD (prior to May this year, anyway). It may be worthwhile having a read through the \'README - crashes and other problems\' post in my signature to see if there is anything which can help.
http://climateapps2.oucs.ox.ac.uk/cpdnboinc/results.php?hostid=447969

Note that deadlines are ignored by the climate project (you can finish a model years past the deadline, and the project is happy), but having a task in deadline trouble will disrupt scheduling on your Boinc client.

---

I'm a volunteer and my views are my own.
News and Announcements and FAQ

Is there a way to enforce the resources sharing AS I set it?

Resource share is not a short term matter; it applies over months.
Your climate model may well be run for a week or so, (depends on how slow your computer is), but then BOINC will work out that it can get it done in the time allowed, and will start to \"pay back\" the excessive time by running your other projects WITHOUT the climate model. (For a while.)

Unless of course, you continue to manually fiddle with your work units; in which case BOINC will never able to work out a suitable schedule for itself.

Don\'t panic! Leave that to BOINC! (That\'s what \"High priority\" means. :) )

Thanks to both of you.
Current deadline for 2008/05/12.
Set to HadSM3 for now, until I get a faster machine.

---

I have the opposite problem to this user:

I want climate predictor to run all the time on 1 core of my quad systems, but it does not run at all! This is because the other projects are cheating with the time settings - I sence an interproject quabble here. CPDN sets the deadline to about a year in the future, while the other projects I am connected to set theirs to a few days (often 2 or 3 days). As a result they are always in the \"Running High Priority\" Status, and CPDN does not get a look in as the next one goes straight to that status and these projects keep picking up more work.

Malariacontrol seems to be the worst offender, but Tanpaku, Boincsimap and World Community Grid also do this.

HOW Can I Adapt the Settings so that CPDN actually runs once and a while???

Thank you,

Whenever Climate Predictor start to run, it prevent ANY other projects from running. It always run as \"High priority\". It completely ignore any resources sharing preferences.

Is there a way to enforce the resources sharing AS I set it?

If left unchecked, Climate Predictor will run for months on end, without ever stoping to let any other project any chance to run.
I need to manualy suspend Climate Predictor if I want other projects to do any work.

---

The answer is the same as for the previous poster - leave BOINC to get on with it and it will.
If you keep changing things yourself, BOINC has to keep re-adjusting it\'s schedule to try and meet your new requirements. And it takes time for BOINC to work things out each time.
Each time you add a new project, BOINC has to re-adjust.

Try setting all of your projects to the same amount of Resource share to start with, until BOINC switchs between them regularly, then change 1 or 2 if needed.

---

Backups: Here

I was just asking for help in contributing to the project. I thought this was a help forum.

\"The answer is the same as for the previous poster\". Does this mean that you cant be bothered thinking about it or telling me the answer?

\"leave BOINC to get on with it and it will\". Get on with not doing what I want it too in the same way it has been doing. Why would I want to do that?

\"it takes time for BOINC to work things out each time\". What? It\'s a computer program! Thats what it does.

If thats your effort at help I\'ll work it out myself.

The answer is the same as for the previous poster - leave BOINC to get on with it and it will.
If you keep changing things yourself, BOINC has to keep re-adjusting it\'s schedule to try and meet your new requirements. And it takes time for BOINC to work things out each time.
Each time you add a new project, BOINC has to re-adjust.

Try setting all of your projects to the same amount of Resource share to start with, until BOINC switchs between them regularly, then change 1 or 2 if needed.

If thats your effort at help I\'ll work it out myself.

In the end, that\'s about all that can be done with this sort of boinc-related problem. (You\'d have to search quite a while to find someone more frustrated with boinc\'s resource management algorithm than me.)

BOINC manages for the long term and will balance time for your chosen projects according to your choices for proportions alloted each project. In doing so, it takes account of deadlines, ...

Given that boinc looks at balancing things over the long term, your short-term druthers might get trampled along the way. (That\'s my gripe, there\'s no easy way to make individual machines in a group do what one wants.) One can set preferences but those preferences can become lost in the fog of boinc trying to be all thing to all participants, and all one\'s boxes.

Les was simply advising you that, for typical participants, boinc should be allowed to work out the long-run according to your specified preferences. That should work for most people, especially those with single machines with single Operating Systems. It breaks down when we have multiple dual-boot boxes and want different priorities for Windoze and Linux, different mixes of CPDN Model types, regular vs. Beta, SAP, ...

Given that you\'ve been attached to the Project for more than a year, you\'ve had plenty of time explore the Boards/Wiki, et al, to learn how things work and realize that your \"Does this mean that you cant [sic] be bothered\" comment is way off base.

By the way, it isn\'t the best way to manage your box, but you can set affinity for CPDN to occupy a particular core.

---

"We have met the enemy and he is us." -- Pogo
Greetings from coastal Washington state, the scenic US Pacific Northwest.

I want climate predictor to run all the time on 1 core of my quad systems, but it does not run at all!

The best way to try and get CPDN running on one core is to arrange your resource shares so that CPDN is 25% of the total (or even a bit more). For example, if you\'re running 7 projects at resource share of 100 try bumping CPDN up to 200.

Malariacontrol seems to be the worst offender, but Tanpaku, Boincsimap and World Community Grid also do this.

I\'d agree on MCDN, can\'t comment on the others as I don\'t run them.

I find that BOINC regularly preempts CPDN between checkpoints on my laptop when it decides that a newly downloaded MCDN task has to run immediately (resource shares set to 100 for CPDN and 25 for MCDN). It\'s particularly nasty because I run with network disabled overnight, which allows MCDN to build up a long term debt. BOINC always works that off by downloading one MCDN task and running it to completion. At which point CPDN is rescheduled but only runs for a couple of minutes at most because BOINC decides it needs more work from MCDN to work off the rest of the LTD! It\'s a good job I\'ve always had my preferences set to keep applications in memory ...

---

"The ultimate test of a moral society is the kind of world that it leaves to its children." - Dietrich Bonhoeffer

Hi Skgiven

I\'ve had a look at the details of your computers and what climate models you have on them:

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/hosts_user.php?userid=189277

You seem to currently have two quads and a dual-core. The dual has one model and the quads have I think 5 and 3 models. All the models are slabs. (There\'s a 4th computer that hasn\'t done CPDN for many months.)

These slabs running 24/7 with no competition from other projects should finish in less than 20 days each on these computers. Their deadlines are about 50 weeks from download. I would become rather frustrated if I still had one of these models for example going into phase 3 nine months after I\'d downloaded it. I\'d become tired of so many models, none making fast progress.

If these computers were mine I\'d choose a radically different solution. Out of the 10 available cores I\'d decide how many in total I wanted to devote to CPDN. I\'d choose a number that coincides with what\'s available on whole computer(s) eg 2, 4 or 6. I\'d then decide which computers to devote only to CPDN in the long term. Eg one quad, the dual, or quad + dual. I\'d let the extra one or two computers only crunch non-CPDN projects.

In order to implement this plan more quickly I would be prepared to set some projects to No new work and if necessary move complete boinc folders from one computer to another. This would avoid abandoning or aborting models.

You\'d need to plan what you want to do but the amount of work and time involved wouldn\'t be much. After that you wouldn\'t need to think about the boinc scheduler and you\'d have the satisfaction of completing your climate models quickly, while still crunching other projects.

If you\'re interested in this sort of long-term solution, members here can help you do it.

---

Cpdn news