Pages: [1]
[BOINCstats] skivelitis
 
Tester - BOINCstats SOFA member
BAM!ID: 55452
Joined: 2008-07-13
Posts: 559
Credits: 179,391,028
World-rank: 5,219

2010-10-31 17:52:12

Found something "interesting" while browsing the BOINC forums: http://boinc.berkeley.edu/trac/wiki/ClientSchedOctTen Pay particular attention to the note under the section Proposal: credit-driven scheduling. Does anyone feel this is a good idea?



Saenger
 
Tester - Translator
BAM!ID: 5
Joined: 2006-01-10
Posts: 1735
Credits: 107,003,843
World-rank: 7,480

2010-10-31 19:06:32

skivelitis wrote:
Found something "interesting" while browsing the BOINC forums: http://boinc.berkeley.edu/trac/wiki/ClientSchedOctTen Pay particular attention to the note under the section Proposal: credit-driven scheduling. Does anyone feel this is a good idea?

Yes, definitely.
It's being discussed in the developers mailing list currently.
Grüße vom Sänger
STE\/E
Tester
BAM!ID: 57534
Joined: 2008-08-27
Posts: 2078
Credits: 1,910,547,903,558
World-rank: 2

2010-10-31 21:56:58

There's nothing wrong with it because as it state's "assuming the host is attached to multiple projects" all you have to do is run the Projects the way I have since BOINC started, only run 1 at a time & the Project is forced to give you work if any is available.

If in the event DA figures out a way to not give you work unless you run other or Multiple Projects then it will be time for every one to stop running BOINC and just run the regular DC Projects.
Crystal Pellet
 
Tester - BOINCstats SOFA member
BAM!ID: 64136
Joined: 2009-01-12
Posts: 4517
Credits: 355,394,684
World-rank: 3,201

2010-10-31 22:04:04
last modified: 2010-10-31 22:05:56

I can't say that the idea is good, whilst it's green and small like an early June apple (Northern Hemisphere).

Most crunchers are expecting that the resource share is based on cpu-runtime and not on credits/RAC.
If the purpose is to penalize the projects who give away high credits (the proposal looks like this) and therefore lower granting projects would get more cputime, one could reach the opposite effect. Crunchers will lower the resource share or more likely skip the lower granting projects totally.
It looks like the developers want to change the behaviour of the users.
IMHO the developers should develop a good, stable programme and ask the users what could be improved.

Example: Make out of 1 PC with different cards more devices/hosts with their own work fetch policy.
Keep it simple: use only wallclock time as a reference for scheduling and fetch policy.

Running High Priority still behaves very strange: Later received tasks with a later deadline run before the first received ones (all of the same project). Repair that first.
BoincWorld: Stop comparing apples and oranges and don't throw all that fruit on one combined Boinc-bunch. After some time it begins to smell
Combined stats are worthless as all credits are worthless. Only the work done for science is valuable.

My 2½ cents.

sorceress
 
BAM!ID: 12734
Joined: 2006-11-19
Posts: 35
Credits: 5,279,760
World-rank: 68,398

2010-11-02 15:34:56

Anytime I hear about D.A. being involved with credit issues, I want to throw things.
D.A. needs to keep his dang nose out of the credit department and concentrate on
making BOINC a more stable, efficient DC platform. I will not waste my computers crunching
low-paying projects. Each computer cost me $11USD to run each month. Good credits
levels are the only compensation I have to justify that expense.

If D.A. continues to screw with the credit bases, I will stop participating in DC computing
and save that money. Projects are the only ones who have the right to determine how much
credit they issue, NOT David Anderson!! That's MY $.02!!
noderaser
 
BAM!ID: 13859
Joined: 2006-12-03
Posts: 827
Credits: 169,222,273
World-rank: 5,423

2010-11-03 03:15:49

To be fair, I think it would be better if the credits were more equal for the contribution given across the projects. While it is fun to compete with credits and the like, the true spirit and purpose of BOINC is to help out organizations that don't have access to huge computing resources. It's not fair to the projects if they have to keep readjusting their credit granting methodology just to keep people crunching for them. Especially when people can be downright militant and nasty regarding credit issues when dealing with the project administrators.

As for the cost of running computers 24/7 for BOINC, I stopped doing that when I bought my own house and no longer had utilities included in the rent. Now, I only have one computer that is on 24/7, my DVR--and the rest are in sleep mode/turned off unless I'm using them. It is a rather odd sensation, as my office is much quieter at night without the hum of many cooling fans.
Konrad Strafer
 
BAM!ID: 381
Joined: 2006-05-14
Posts: 39
Credits: 72,743,819
World-rank: 9,828

2010-11-03 18:40:46

I think the solution is to have the credits built into BOINC, so that no one project can adjust credits higher than others. The issue would go away if BOINC itself fixed the credits independently of the control of the project. While it is true that idealistically the science is the motivation, most crunchers like to see their credits go up as fast as possible. That's why some users will switch to a project that pays more credits. It's just normal human psychology.
Pages: [1]

Index :: BOINC :: D.A. at it again?
Reason: