Message boards : Number crunching : How do I get a second model?
Message board moderation
Author | Message |
---|---|
Send message Joined: 26 Aug 04 Posts: 14 Credit: 123,062 RAC: 0 |
I have two P4 3.0 HT computers that connect via dial-up and I was wondering how I can get a second model to download on each so when it completes other workunits for other projects, I don't have the processor sitting half idle till I can connect again. I'm running BOINC 4.25 Thanks |
Send message Joined: 5 Aug 04 Posts: 1283 Credit: 15,824,334 RAC: 0 |
You'll need to modify the network connection period in your BOINC general preferences. Set 'Connect to network about every' greater than the estimated number of days to complete your current CPDN model (the maximum value is 10 days). But you need to be aware that changing this setting will affect all projects your hosts are connected to as soon as you force an update for CPDN. The scheduler response will contain the modified preferences, and these will automatically be sent in scheduler requests to any other projects you're connected to. All projects will end up downloading more work, possibly giving you deadline problems. I suppose you could prevent your other projects from downloading more work by setting your firewall to block communications with the other projects until you've downloaded another CPDN workunit and reset the network connection parameter. "The ultimate test of a moral society is the kind of world that it leaves to its children." - Dietrich Bonhoeffer |
Send message Joined: 31 Aug 04 Posts: 239 Credit: 2,933,299 RAC: 0 |
I don't think you can without making the queue real long to get one, but then you will be flooded with work for the other projects. This is a feature we have asked for, but at the moment it is not considered high priority. If you are concerned about not keeping busy you should try to get into all of the open projects ... Since I joined all 5 (yes I was lucky to get into all of them) I have only come close to running out of work once. So, that is the bad news ... |
Send message Joined: 26 Aug 04 Posts: 14 Credit: 123,062 RAC: 0 |
So suspending the other projects in 4.25 won't force the running project to pull a second model for the idle "CPU"? |
Send message Joined: 31 Aug 04 Posts: 239 Credit: 2,933,299 RAC: 0 |
> So suspending the other projects in 4.25 won't force the running project to > pull a second model for the idle "CPU"? I don't think so... be the first to try it ... :) |
Send message Joined: 26 Aug 04 Posts: 14 Credit: 123,062 RAC: 0 |
> > So suspending the other projects in 4.25 won't force the running project > to > > pull a second model for the idle "CPU"? > > > I don't think so... be the first to try it ... :) > Nope, doesn't work :-) |
Send message Joined: 31 Aug 04 Posts: 239 Credit: 2,933,299 RAC: 0 |
> > I don't think so... be the first to try it ... :) > > Nope, doesn't work :-) The systems are getting smarter, but they are still not as flexible as we would like. I agree, I would prefer to have one model for each processor node, mostly as a safety valve so that I did not run out of work. But we can't make that a preference setting, yet. Heck, I would prefer that once I start a particular WU on a processor (with the exception of CPDN) it would run that WU to completion rather than to have 4-5 partly done ... I mean for most projects, they finish in a reasonable amount of time. Even better, I would like BOINC to not stop processing the model when it is 5 minutes from completion ... |
Send message Joined: 24 Sep 04 Posts: 1 Credit: 398,243 RAC: 0 |
I run P4 3GHz HT for CPDN. At one time, I set CPDN to have more than 50% share in total and BOINC automatically downloaded a further WU. I was >300 hours away from the end of the original WU, so it was nothing to do with my cache settings. After thinking anout it for a while, I assumed that it did this because the second HT thread demanded CPDN work that could not be satisfied by the original WU. So, try setting CPDN >50% share in total and see what happens. You may need to maunually update the project to force BOINC to download. Alternatively, assuming you are running BOINC 4.25, suspend the CPDN result (not the project). This should download a new WU because BOINC needs to satisfy the request for CPDN work. (Once again a manual update may be needed to kick the client into a response). Once the new WU is downloaded, resume the original result. Good luck! Richard |
Send message Joined: 1 Sep 04 Posts: 2 Credit: 297,724 RAC: 0 |
I run a P4 3.0 ghz running nothing but cpdn. It dose two models at one time and has never stopped computing since september 1 2004. This seems to be the best way to run cpdn running other projects with it seems to create problems with the ram. <img src='http://www.boincsynergy.com/images/stats/comb-1101.jpg'> |
©2024 cpdn.org