| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Linux & XP differences |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 23
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello urban peasant,
Nice to hear from someone who is moderate and rational. We are giving 3 types of scores - CPU hours contributed (which BOINC measures much more accurately than the UD agent); results returned (which might be invalid and can vary greatly in difficulty between projects and somewhat within projects); and points, which offer the hope of normalizing the effort contributed per result between projects. However, when closely examined, points fail to normalize the differences between compilers for BOINC projects and have other problems with the UD agent formula. It would be easy to fairly award points for a single project using a single compiler. Trying to accommodate all the variations is where the problem comes in. So then the question arises - well, what is satisfactory to as many people as possible if nobody can think of a solution that is both fair and practical? Some people argue for a system which they claim is better (or even perfect!). Some want another fudge factor added to the scoring formula while admitting imperfection. Some just don't care.  The main problem is that any serious discussion seems to quickly degenerate into mere name-calling, which means that most people just decide to avoid the topic, even if they have some ideas or opinions of other ideas. I would be willing to adopt a different scale ratio between BOINC and WCG points for each OS / project, even though I can see some problems this would not solve, but I don't see any possibility of consensus in the flame wars on the subject, so why waste time trying to think of admittedly imperfect solutions.  Tiredly, Lawrence |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
There have been some really good, productive discussions on how to implement a fair points system in the past. I won't rehash the arguments here, but they can be found by searching the archives.
The state of the BOINC discussion on credit is summarised in the proceedings of the 2nd BOINC Workshop at the University of Geneva: http://boinc.berkeley.edu/ws_06/credit.txt |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hate to be a sceptic, but that document makes it seem that Dr Anderson and company are more interested in "Socially Engineering" ie conning us into liking the present system, or a variation upon it, rather than actually fixing the credit system.
Only one vauge seeming proposal is mention under fixing the system, while there are 4 proposals under "social engineering". |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi 5 and a half of 13,
There has been a lot more palaver since the conference. Currently, Dr. Anderson is scheduling a quick fix for Linux with the upcoming BOINC 5.8. There is talk about more to do later, but that is ongoing and subject to modification as the discussion proceeds. I am not following this topic but I expect that there is some mention on the developer's forum. Lawrence |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
|
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
One of the things read was, that the BOINC benchmark for Linux runs on a priority of 'Nice 19'. I don't know that OS at all, but presume it to be something similar to the 'low(est)' priority in Windows giving away to anything equal or higher. It was proposed to change the 'Nice' for the Core Client (BOINC.exe) that runs the benchmark to a low number, so at least the benchmark does not get affected if it so happens, that you were in the middle of a heavy photo editing session.
----------------------------------------Of course U could always get up one night at 4 AM shut down all the apps and services and force the benchmark and go back to bed. Than, it would run from thereon every 5x24 hours at 4 AM, thus ensuring the benchmark is least effected by your daytime chores. This way, u have the best 'legal' benchmark with a stock BOINC, without editing files and return to bed, sleeping like a rose. It does not solve the disparity in benchmark compiling between the OSses, but at least u get it maximized without hacking until 5.8 arrives. Whilst, be patient as the proposed intermediate WCG specific solution is being compiled and field tested....i think the Linux folk are in for a fair treat....all OSses made equal on the BOINC points awards.... i day dream coming true :o cheers
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 1 times, last edit by Sekerob at Nov 4, 2006 7:24:49 AM] |
|||
|
||||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Thanks for the updates, good to know there is more going on behind the scenes.
|
||
|
|
Jake1402
Senior Cruncher USA Joined: Dec 30, 2005 Post Count: 181 Status: Offline Project Badges: 
|
urban peasant has it right. Crunch for the reason, not the reward.
----------------------------------------I think the scientists have better things to do than points disparities.
Join the Chicago-IL-USA team!
2 AMD FX 8320/AMD R9 270X/Win 10 2 AMD FX 8320/AMD RX 560/Linux Mint 20.3 (both computers DOA) Intel Pentium G240/Win 10 |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
hi all....
As has been mentioned, a new linux client (5.8) is supposed to overcome the points disadvantage of linux https://secure.worldcommunitygrid.org/forums/wcg/viewthread?thread=9545 I see on the Boinc download site that 5.7.5 is still unstable and for testing only.... so how much further away is 5.8 going to be? Rob |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
It is very close. There hasn't been an update to the alpha version for a little while now, and David Anderson has expressed the hope that 5.8 will be released in the near future (it is feature complete, I think).
|
||
|
|
|