I have a problem with the size of these workunits

Message boards : Number crunching : I have a problem with the size of these workunits

To post messages, you must log in.

AuthorMessage
ssoxcub@yahoo.com
Avatar

Send message
Joined: 8 Jan 12
Posts: 17
Credit: 503,947
RAC: 0
Message 76531 - Posted: 20 Mar 2014, 8:07:41 UTC

I can choose between 1 and 24 hours what does that mean?
ID: 76531 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 76544 - Posted: 23 Mar 2014, 17:10:48 UTC
Last modified: 23 Mar 2014, 17:15:15 UTC

The default for the preferred runtime is 3hrs. For a given protein and set of rules to apply to studying it, your machine basically computes as many models as it can within in the preferred runtime. So the amount of data downloaded per task is the same, regardless of how long you crunch on it. The amount of data uploaded with your results will be larger depending on the number of models your machine completes. The runtime of the next model is not entirely predictable and so this is a "preference", not a "rule". If the last model happens to take longer than expected, the task will be marked as completed and reported back once it completes (there are also "watchdog" functions that will report it back if it does not seem to be making progress in your environment).

So to give your some flexibility about how the tasks run, this setting can help you set things up to better assure you complete tasks before their deadlines, share your machine with other BOINC projects you may be running and etc.

If your machine primarily crunches only Rosetta, or is running more than 10 hours per day, I often recommend selecting the 24hr runtime. This reduces the number of hits to the project servers, and lets your machine run for the longest time with a single set of downloaded files.

Beware, any changes to the runtime preference take place immediately and will effect your currently downloaded tasks (although BOINC Manager will take a while to learn how to properly estimate the remaining runtime). So, I suggest you make changes to the value when your machine has a small cache of work, and only changing the value incrementally. For example, if you want to go from the 3hr default, to the 24hr runtime preference, I'd suggest doing so over the course of a week in 3 or 4 steps. This just helps ensure you don't get a pile of tasks and then have them run longer than BOINC's estimates, and thus you don't complete them all before their deadlines. Basically, when you have a task you have not starting computing against, it's estimated runtime should be closer to your current runtime preference than to your prior runtime preference, before you bump the preference to a higher level.

With that basic outline, please post any other questions about the specific situation you might like to address.
Rosetta Moderator: Mod.Sense
ID: 76544 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote

Message boards : Number crunching : I have a problem with the size of these workunits



©2024 University of Washington
https://www.bakerlab.org