Message boards : Number crunching : Taking a computer offline
Message board moderation
Author | Message |
---|---|
Send message Joined: 17 Feb 09 Posts: 31 Credit: 1,505,895 RAC: 529 |
Im taking one of my computers offline to add a graphics card and use as a golf simulater. The new location will not allow constant internet hookup and it will have to be manually moved to update the trickles/credits/zip files/new work units etc. My question is how long can I go without it connected and still succesfully run cpdn projects? Rick |
Send message Joined: 29 Sep 04 Posts: 2363 Credit: 14,611,758 RAC: 0 |
I believe the server considers a task abandoned after receiving no trickle for six weeks. It might then conceivably send another fresh task from the same WU to another member. But when your task did eventually trickle the server would still accept everything. One member who works on a submarine took his tasks onboard for a long stint. He had no internet on the sea bed. A BBC member with a 160-year model sent its first trickle more than two years after downloading it. Needless to say the model was never completed because the BBC project closed down before the next trickle was ready, but the server accepted that first trickle. I\'d advise a connection about every 5 weeks if you can do it. You\'d have to make sure the computer has plenty of work in advance to get on with. If this is likely to be a problem let us know because there may still be a way to edit a file to trick the server into thinking the machine has more cores than in reality. This makes the server send extra work, way above usual. (Not advisable of course for most circumstances.) Cpdn news |
Send message Joined: 17 Feb 09 Posts: 31 Credit: 1,505,895 RAC: 529 |
I\'d advise a connection about every 5 weeks if you can do it. Ive set the preferences on that host to 10 days yesterday got one additional work unit and currently have 3 x hadsm3 work units. One at 57%, one at 5% and one at 0%. Total projected hours ~ 840 hrs. With 2 cores I will be out of work for one of the cores in ~ 485 hrs (~20 Days). How do we go about increasing the downloads so I can get about 30 - 45 days work? I should be able to connect it monthly to keep things current. Rick |
Send message Joined: 3 Oct 06 Posts: 43 Credit: 8,017,057 RAC: 0 |
You could create a cc_config.xml file, and change the <ncpus>N</ncpus>-setting. This can be used to fool the client into thinking your computer has more cores than actually available. This setting is intended for testing and, I think, will result in N models running. When you have downloaded enough models I, therefore recommend changing this setting back or deleting/renaming the cc_config.xml file. For more info on this config file, click this link: cc_config.xml |
Send message Joined: 17 Feb 09 Posts: 31 Credit: 1,505,895 RAC: 529 |
You could create a cc_config.xml file, and change the <ncpus>N</ncpus>-setting. This can be used to fool the client into thinking your computer has more cores than actually available. This setting is intended for testing and, I think, will result in N models running. When you have downloaded enough models I, therefore recommend changing this setting back or deleting/renaming the cc_config.xml file. For more info on this config file, click this link: cc_config.xml I read that info and to be honest it is mostly foreign to me. I have manually added xml files to my seti folder in the past but to actually write one is a bit beyond my ability at this point. Any other leads to read? Rick |
Send message Joined: 3 Oct 06 Posts: 43 Credit: 8,017,057 RAC: 0 |
If you do not feel comfortable making such a file, I would recommend not going down this particular path. I can\'t think of an alternative, myself. |
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
The hadsm3 M-H models are 4 phase, so they take 1/3 longer then the 3 phase hadsm3\'s. Set your prefs to only get these. WARNING: hadsm3 multi-phase models do NOT like being interrupted at phase change, when they\'re busy converting lots of files from one format to another, and adding them to a zip file. Interruptions cause them to rewind, often to the start of the model. This includes automatic project switching by BOINC, for computers running multiple projects. Backups: Here |
Send message Joined: 29 Sep 04 Posts: 2363 Credit: 14,611,758 RAC: 0 |
Here\'s a possible alternative. You could return on that computer to Boinc v. 5.2. (You probably wouldn\'t be able to see the model graphics.) On these early versions if you suspend tasks, the server sends more. Suspend the new tasks, get yet more. This doesn\'t happen on recent Boinc versions. In fact now if you suspend a task the server for that project refuses to send anything at all. Old versions are available here. If you do want to revert to Boinc 5.2 it would be best to set CPDN to No new work, run all tasks to completion, upload them and them make sure they\'ve reported. Then when there\'s no work left, uninstall Boinc and get the old version, installing from scratch. You could instead fetch work from other projects to keep the computer busy but I can\'t think of projects with long enough deadlines. Cpdn news |
©2024 cpdn.org