credits and workunits: a fair system needed for Boinc

Message boards : Number crunching : credits and workunits: a fair system needed for Boinc
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3

AuthorMessage
Profile BMH
Avatar

Send message
Joined: 27 May 99
Posts: 419
Credit: 166,294,083
RAC: 125
United Kingdom
Message 1132262 - Posted: 26 Jul 2011, 20:18:05 UTC

I've been a SETI@home participant since '99 but I did stray away for a while while I participated in a non-BOINC-based project. I recently directed my computers back to the SETI because of an issue I had with the points system with the other project which I felt was not beneficial to the science.

For me I enjoy setting myself little targets and comparing participants and computers with similar RAC. I like how we can all offer tips and advice and help out when people run into problems, while at the same time essentially competing against each other.
Brian.
ID: 1132262 · Report as offensive
Profile Mike Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 17 Feb 01
Posts: 34253
Credit: 79,922,639
RAC: 80
Germany
Message 1132269 - Posted: 26 Jul 2011, 20:32:15 UTC


If i would agree on a change it has to be teraflops.
Easy to compare between projects.



With each crime and every kindness we birth our future.
ID: 1132269 · Report as offensive
1mp0£173
Volunteer tester

Send message
Joined: 3 Apr 99
Posts: 8423
Credit: 356,897
RAC: 0
United States
Message 1132272 - Posted: 26 Jul 2011, 20:34:13 UTC - in response to Message 1132151.  

To put it another way, why should my credits depend on how much work my wingman did?

... because you both did exactly the same amount of work.

Sure, you've got a fancy GPU and spent five minutes doing it, while his six year old Celeron took a week, but the same amount of math was done.

That's equal pay for equal work.

The difficulty lies in differences between processors, differences between the instruction mix in the benchmarks and the science applications, differences in cache sizes and memory speeds.

All of the systems we have are "fair" if by fair you mean "equitable." The credits may not accurately reflect the amount of work, but everyone has the same chance of being overpaid or underpaid, and everyone gets the same.

Benchmarking was an inexact science in the 1950's, when computers were very simple, and you could calculate exactly how many clock cycles it'd take to get through a loop -- no cache faults, no wait states.

It's even harder to accurately measure work now.
ID: 1132272 · Report as offensive
Profile Gary Charpentier Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 25 Dec 00
Posts: 30608
Credit: 53,134,872
RAC: 32
United States
Message 1132295 - Posted: 26 Jul 2011, 21:16:55 UTC - in response to Message 1132272.  

The difficulty lies in differences between processors, differences between the instruction mix in the benchmarks and the science applications, differences in cache sizes and memory speeds.

So should the benchmarks be required to be optimized if you run an optimized application?

ID: 1132295 · Report as offensive
1mp0£173
Volunteer tester

Send message
Joined: 3 Apr 99
Posts: 8423
Credit: 356,897
RAC: 0
United States
Message 1132323 - Posted: 26 Jul 2011, 22:10:41 UTC - in response to Message 1132295.  

The difficulty lies in differences between processors, differences between the instruction mix in the benchmarks and the science applications, differences in cache sizes and memory speeds.

So should the benchmarks be required to be optimized if you run an optimized application?

That is actually a great question, and one that was discussed at great length some time ago.

The consensus was "no" -- if you want to run an optimized app., you should get some advantage to go with the hassles.

I haven't followed the latest, but the previous credit system attempted to approximate benchmark * time by counting passes through loops and estimating the work done each time.

The problem is the need for some conversion factor that was to be determined experimentally to get you to the proper credit.

If you're a project scientist, that scaling factor has nothing to do with science at all. So, too many projects have just taken a swipe at it, and if they overpaid, no big deal. They can double the amount of "nothings" given per work unit and it's all the same to them.

If anything, it is in a projects' best interest to overpay.

Last I knew, most projects (including SETI) overpaid by about 3:1.
ID: 1132323 · Report as offensive
tbret
Volunteer tester
Avatar

Send message
Joined: 28 May 99
Posts: 3380
Credit: 296,162,071
RAC: 40
United States
Message 1132426 - Posted: 27 Jul 2011, 4:39:35 UTC
Last modified: 27 Jul 2011, 5:10:29 UTC

As far as I'm concerned, we could do away with numbers entirely for scoring SETI@Home.

We need scores like, "You really crunched the crap out of that one;" or "Hey! Not even StanQPublic's 590 crunched that unit as fast as you did!" or "Maaaan, are you crunching on an Atari? 240 hours? You're $%*%in' me, right?"

You see, I just want to crunch.

I wanna get integers between my teeth and my points to be floating on a sea of FFT fire.

Then we can all get in a big room and jump up and down shouting "crunch! crunch! crunch!"

I want electrons appearing at THREE places at once because I'm running three WUs simultaneously. I wanna CRUNCH.

I'm thinking BIG credit: I want SETI@Home to hasten global warming so we can see Antartica (the land mass) so we have more places to put more computers with more GPUS for crunching more SETI WUs. Our GPU fans should be causing tornadoes, hurricanes, a change in the tilt of the Earth! Water cooling should be no problem for anyone near a coastline.

I'm not talking this small-fry credit-stuff; "Did I get 100 credits while you got 101?" Nooooo, I mean I want to meet the SETI@Home event horizon where it becomes economically impossible for any company to do anything other than build ever-faster GPUs for SETI crunching.

All I need for comparative purposes is essentially what we have now with RAC; but rather than actually calculating anything, we could just use a "standard question" and a "Magic 8 Ball" to generate information just as useful.

"Oh SETI@Home, am I crunching efficiently today?"

"Reply hazy, try again."

"Oh SETI@Home, am I crunching efficiently today?"

"My sources say no."

And then I want to run it all down a 100Mbit line so I can watch the cable burn like a .25Amp fast-blow fuse struck by Technicolored Quasar ultra-lightning.

And when we're done and the first alien signal is found, we can all -in three part harmonical convergence- go somewhere and count the work unit credits together like a Skeksis mantra:

"Oooooooooooooooone."

EDIT: I got it wrong. It wasn't the Skeksis, it was the Mystics.
ID: 1132426 · Report as offensive
Profile Bill Walker
Avatar

Send message
Joined: 4 Sep 99
Posts: 3868
Credit: 2,697,267
RAC: 0
Canada
Message 1132515 - Posted: 27 Jul 2011, 11:38:36 UTC - in response to Message 1132426.  

What he said.

ID: 1132515 · Report as offensive
-BeNt-
Avatar

Send message
Joined: 17 Oct 99
Posts: 1234
Credit: 10,116,112
RAC: 0
United States
Message 1132530 - Posted: 27 Jul 2011, 12:34:03 UTC

I think somebody forgot their medicine. However I do agree.
Traveling through space at ~67,000mph!
ID: 1132530 · Report as offensive
Previous · 1 · 2 · 3

Message boards : Number crunching : credits and workunits: a fair system needed for Boinc


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.