I'll try and answer a few of them...
cesium_133 wrote:
I looked through all the Q&A topics, read the manifesto for Rosetta, and looked at other projects' specs.
The questions I have are scientific in nature and go to the heart of Rosetta, primarily. They go to some more basic, yet mostly unspoken-on, issues regarding basic procedures within BOINC... so here I go...
1. Most of what I run is Rosetta, and I have a big load of projects in that queue... with 1 each from 6 of the other 10 I have signed up for. They're at the bottom of the queue, and I have received no further projects lining up after them. Should I expect to see more Rosetta after these others? I am allowing equal usage time for all 11 projects, and I suspect the other ones just don't have as much work to do. How might I see future things line up? I want to use my PC for something all the time...
You'll get more Rosetta. If there is no work for a particular project BOINC will simply keep getting work from whatever other projects you have running. BOINC determines this based on short and long term debts. A quick summary would be that short term determines which wu to crunch of the ones already downloaded - it rotates projects based on your settings and wu deadlines. Long term debt (LTD) determines which project to ask for work from next. Check out the Un-official BOINC Wiki link =
http://www.boinc-wiki.info/Short_Term_Debt#Short_term_debt for a quick description.
A more detailed explanation of LTD would be that every time you get work from a project it decreases that projects LTD value and increases your other projects LTD value, with the sum of all projects debt being equal to zero. BOINC will not download work from a project that has a negative LTD unless there is no work available from any other active project (idle CPU or space left in work queue). Lets look at a little example of this based on a box crunching 2 projects with equal weight - one with lots of work (P1), and one with sporadic work (P2).
1. P1 downloads work so it's LTD decreases; let's say it drops to -10 just to make it easy
2. Since P1+P2 = 0 we can figure out pretty easily that P2s LTD would now be increased to +10 to make the total value = 0
3. The next time BOINC tries to get work it will try from P2 as P1 has a negative LTD
4. P2 only has sporadic work, so none is available - BOINC contacts P1, gets more work and changes LTDs appropriately (P1 = -20, P2 = 20)
5. This continues for several iterations until P1=-1000 and P2 =1000, with BOINC continuing to try to get work from P2 prior to trying P1
6. Eventually P2 gets work and BOINC contacts it and downloads some (P1=-990, P2=990)
7. Since P1 still has a negative LTD, BOINC will continue to fill it's cache with work from P2 until P2 runs out of work or the LTD evens out.
At least that's the theory...
cesium_133 wrote:
2. In Rosetta, the work units have some funny names and numbers attached... ab initio norelax homfrag this and that (from the beginning...); IGNORE THE REST; SAVE ALL OUT; jump pos; and such. Anyone know what this terminology means? Can't find it anywhere...
I can't help you here - try posting on their forums for more info regarding wu naming.
cesium_133 wrote:
3. Is BOINC a more-or-less permanent thing, soliciting or allowing itself to be solicited for new projects for the foreseeable future?
BOINC server software is free to download and use so it'll be around as long as people find the DC framework it provides useful.
cesium_133 wrote:
4. In the Rosetta crosshairs, lower right hand corner, I sometimes see a red dot or 2, rather than a green. Any idea what that is? Also, I have seen low accepted energy and low RMS error change without regard to realtime figures... any idea why?
Can't help you here either - try Rosetta's forums!
cesium_133 wrote:
5. Should I assume that the projects w/o work units (whichever they are) are dormant, finished, being analyzed, or what? And should I keep them in my list of ones to contribute to? I picked very carefully when I was deciding which ones to take on... someone else can have prime #'s and SETI and chess permutations...
For the most part projects make good use of the data (at least I'd like to believe that!). However, check the individual project forums for complaints and issues. If you want to be sure your contributions are for the "greater good", check to ensure the project staff are publishing their data and are on top of the forums. If the project has been around for several years and there have been no meaningful publications or linked usages (e.g. Simap maintains a public database as their "publication") coming out of the lab then who can really say what the data is being used for. Another common gripe people have with some projects are the number of results needed for a quorum - several projects will require 3+ identical successful results per wu which some feel is a waste of CPU processing power when 1 or 2 successful results would have been adequate for their purposes. Just do your due diligence and check the forums and you should be alright. When in doubt you can always check on BOINCStats popularity page -
http://boincstats.com/page/project_ranking.php - it's user generated so is a somewhat good measure of project stability/usefulness.
cesium_133 wrote:
6. If no work comes my way at any time, how might I request a load?
If the project has any work available for your platform you can always open BOINC manager (in advanced mode), select the project, and hit "Update" to force a scheduler request. Most sites have a status page link on their front page to let you know if work is available.
cesium_133 wrote:
Thanks for your help, guys, and I hope someone with more insight than I can answer these. They're good ?ions, and can be answered by someone in the know faster than I can (try) to research them; I'm not a big RTFM guy anyhow.
Hopefully I've helped a bit...
Login
Register
Portal
Search
