Akos Optimizations require installing Calibrating Clients 2 !!!

UBT - PaulT
UBT - PaulT
Joined: 22 Jan 05
Posts: 12
Credit: 86,678
RAC: 0

RE: I believe in keeping my

Message 32031 in response to message 32024

Quote:
I believe in keeping my system as vanilla as possible so I will use BOINC as offered by UCB. When Akos offers a science app that is 4x faster than what the project offers how can I ignore it?

So you DON'T believe in keeping your system as vanilla as possible then.

If you did you would use the science app offered by the project.

Erik
Erik
Joined: 14 Feb 06
Posts: 2,815
Credit: 2,645,600
RAC: 0

RE: RE: I believe in

Message 32032 in response to message 32031

Quote:
Quote:
I believe in keeping my system as vanilla as possible so I will use BOINC as offered by UCB. When Akos offers a science app that is 4x faster than what the project offers how can I ignore it?

So you DON'T believe in keeping your system as vanilla as possible then.

If you did you would use the science app offered by the project.

I believe he was refering to the structure of BOINC itself. Calibrating clients wouldn't be "vanilla" with their benchmarks. Consider all the people across all projects - how many use something other than "stock". How many installed BOINC, set it going and don't read the forums and so have no clue about tweaked benchmarks. It's not just Einstein, though yes the "optimized" apps do throw credits for a loop. Somehow I don't believe Nothing But Idle Time cares about the credits as much as the actual work, he has no team.

Honza
Honza
Joined: 10 Nov 04
Posts: 136
Credit: 3,332,354
RAC: 0

I'm not much into credit –

I'm not much into credit – CPDN is my #1 where "ill" benchmarks, "smart" calibration or even "fair" FLOPs mean very little...it is the work done that counts! but...

trux (all praise his name) is not continuing on further developments of his greats ideas. Some features are great and still superior to other BOINC cores, some are missing.

SETI went to FLOPs counting. Hopefully, other projects will follow.
Boinc studio core has also credit correction mechanism. It also enables those will to put full CPU power into BOINC by faking cpu number and avoid "Curse of 32".

Not saying a bad word about trux's core, but showing other ways to go as well...

bloed_brot
bloed_brot
Joined: 5 Apr 05
Posts: 70
Credit: 91,124,558
RAC: 0

RE: In your example the low

Message 32034 in response to message 32024

Quote:

In your example the low credit awarded is NOT because of Akos' improved apps, but because the project team throws out the high and low claims and grants the middle one. UWM's approach to granting credit doesn't appear to be equitable.

That said, it is not my responsibility -- or that of any host -- to level the playing field for everyone else. I believe in keeping my system as vanilla as possible so I will use BOINC as offered by UCB. When Akos offers a science app that is 4x faster than what the project offers how can I ignore it?

It is because of Akos improved apps! In my example everybody used Akos optimisations, but do not deploy a boinc version that levels out those shorter than anticipated cpu time thereby corrupting the credit system. Akos are not official, so it is not UWM's job to level out those flaws coming from utilising Akos' stuff. It is YOUR RESPONSIBILITY because YOU CHOOSE TO USE AN UNOFFICIAL VERSION! So you cheat everybody who runs an official version or those that level up to it.

You are so much contradicting yourself that it hurts. You talk of vanilla and then decide to use an unofficial app? Your move lets you claim a quarter only of what was anticipated by the designers of the credit system. You cause this system to fail, you work against the community effort of everybody being treated fairly through the credit system. And all you do is remaining self-centered:

Quote:

The correct solution is at the roots: to get UCB and UWM to implement a proper benchmarking or flops-counting system and to grant credits in an equitable way across the board for everyone. Many people do not have the time or desire to manually install Akos apps or to manually install special clients, both of which implies upgrading them periodically.

So stick to the official client, which does the update automatically for you. But when you decide to use unofficial stuff, the least you can do is to make sure it does not interfer with the official intended way. Once again: By running optmitised unofficial clients, you have to use calibrating BOINC application to prevent you from cheating the community. PERIOD.

Redesigning the credit system is an entire different issue, and regardless of its outcome it has nothing to do with you cheating the community right now.

:
your thoughts - the ways :: the knowledge - your space
:

bloed_brot
bloed_brot
Joined: 5 Apr 05
Posts: 70
Credit: 91,124,558
RAC: 0

RE: I believe he was

Message 32035 in response to message 32032

Quote:
I believe he was refering to the structure of BOINC itself. Calibrating clients wouldn't be "vanilla" with their benchmarks. Consider all the people across all projects - how many use something other than "stock". How many installed BOINC, set it going and don't read the forums and so have no clue about tweaked benchmarks. It's not just Einstein, though yes the "optimized" apps do throw credits for a loop. Somehow I don't believe Nothing But Idle Time cares about the credits as much as the actual work, he has no team.

I hear you, but nobody forces him/her to apply unofficial clients. The mechanism of calibration can be switched on and off for individual projects, too. In the case of Einstein / Albert, you would need to read the forums to get the version. His line of argument is just not cricket.

:
your thoughts - the ways :: the knowledge - your space
:

Nothing But Idle Time
Nothing But Idl...
Joined: 24 Aug 05
Posts: 158
Credit: 289,204
RAC: 0

RE: RE: RE: I believe

Message 32036 in response to message 32032

Quote:
Quote:
Quote:
I believe in keeping my system as vanilla as possible so I will use BOINC as offered by UCB. When Akos offers a science app that is 4x faster than what the project offers how can I ignore it?

So you DON'T believe in keeping your system as vanilla as possible then.

If you did you would use the science app offered by the project.

I believe he was refering to the structure of BOINC itself. Calibrating clients wouldn't be "vanilla" with their benchmarks. Consider all the people across all projects - how many use something other than "stock". How many installed BOINC, set it going and don't read the forums and so have no clue about tweaked benchmarks. It's not just Einstein, though yes the "optimized" apps do throw credits for a loop. Somehow I don't believe Nothing But Idle Time cares about the credits as much as the actual work, he has no team.


Believe me, I'm not callous or indifferent about the credit inequity. I've benefited from it while others have suffered. But you don't achieve equity by asking thousands of hosts to install optimized apps to achieve efficiency, nor to install calibrating clients to compensate for the inadequacies of BOINC or UWM's credit system. The project itself must implement Akos' optimizations for its own sake because it directly benefits. UCB and the projects should provide a proper credit system to assuage the masses who care about it. I wish I could offer a way to make it happen.

Sadir
Sadir
Joined: 9 Oct 05
Posts: 4
Credit: 1,334,663
RAC: 0

RE: Would you like to

Message 32037 in response to message 32030

Quote:

Would you like to prescribe others, what her to do or to leave have? - Quite arrogantly.


I am sorry if my message sound quite arrogantly. I just want to say something in style:
"Please install a callibrating client"
reply: "Please use an optimalized application."
Others CAN use an opt. aplication (not have to):)

Quote:

If you had a seven times faster machine, would you let credit with your arguments the same credit, as if would be seven times slower machine? (No hardware without software.)


Seven time faster machine can do seven time more work.
If I crunch for other project for 1 hour, I got 7 credit. If I crunch for Einstein with opt. app. and callibrating client, I got 30 credit. Should crunching for one project (Einstein) be prefered from crunching from anothers projects? And anothers projects should say "we give them more credit for the work"... and so on, and where will be end of this? More money (credit) for everyone = less value of money (credit).
I only think that good way isn't using a calibrating client but using optimised application as official:)

UBT - PaulT
UBT - PaulT
Joined: 22 Jan 05
Posts: 12
Credit: 86,678
RAC: 0

RE: UCB and the projects

Message 32038 in response to message 32036

Quote:
UCB and the projects should provide a proper credit system to assuage the masses who care about it. I wish I could offer a way to make it happen.

SETI have tried that with (so called) Enhanced and, to be honest, they have made a complete balls up of it. Not only do you get less credit per hour than using the non optimised version of the old science app, you are actually doing MORE WORK per hour. People are up in arms over this and there are some very heated discussions over on their forum.

At the moment, regardless of wether you are using the official app or one of the optimised apps, your computer is still doing the same work, it's just doing it faster. A look at your results shows that you are claiming on average approx 14 credits per workunit, My avreage is approx 45, it will also be about 45 credits for everyone else using the calibrating client.

Look at it this way, you are doing approximatly 4 times the work of someone using the non optimised app, but still getting the same credit. Would you do 4 times the work at your place of employment for the same money you get now?

It also works the other way, some poor sod with one old PC that takes best part of a day to complete 1 workunit is, rightly, expecting his 45-50 credits. Trouble is, his PC is in the same quorum as your computer, and someone else with the optimised app, but still bog standard BOINC client and instead of getting his 50 credits, he gets 14. How fed up is he going to feel?

Before anyone starts the old "It's about the science, not the credit" rubbish, just remember this.

SETI classic introduced Credits, Teams, Tables etc. to bring a bit of competition into the project and to increase the amount of work done for them. If it was not for the credits, most people probably would not have participated as SETI played on the competitive instincts that everyone has.

Matt LO
Matt LO
Joined: 7 Feb 06
Posts: 44
Credit: 386,731
RAC: 0

A suggestion. Perhaps

A suggestion. Perhaps someone (moderator) should place info regarding the calibration clients into the first post of Akos' optimized executables thread, suggesting the installation of one along with a optimized app for fair credit claiming under the current credit system. I like how there's a "how to install" in that thread, and you can put installing a calibration client in the "how to" steps to further drive home the idea, instead of people needing to read 2 threads.

That is, if we have consensus that this is the best thing to do. I, for one, think keeping the credit claims for optimizers on par with the standard application claims per WU is the right thing to do. If you choose to optimize, installing the calibrating client should be part of that process to keep credit relatively fair.

archae86
archae86
Joined: 6 Dec 05
Posts: 3,158
Credit: 7,242,576,951
RAC: 1,309,632

RE: A suggestion. Perhaps

Message 32040 in response to message 32039

Quote:
A suggestion. Perhaps someone (moderator) should place info regarding the calibration clients into the first post of Akos' optimized executables thread, suggesting the installation of one along with a optimized app for fair credit claiming under the current credit system.

Agreed. Then I'd like the moderator to "unsticky" this one. The opening post uses a tone that reminds me of the personal animus which has consumed too much recent discussion on the SETI forums lately.

We had a pretty good optimized clients thread going for a while--if the point is to invite people to review the discussion and reasoning the proposed entry could point to that.

In the spirit of the spendid brevity of the optimized ap thread, perhaps there could be an entry along these lines:

************************

Use of an optimized science ap with the standard BOINC client will lead to very low credit claim per work unit. This may subtract credit from others who happen to share a quorum with one of your results. Many Einstein@Home forum participents have concluded that on balance use of an optimized client--most likely Trux's Calibrating Client--is fairest to all.

A discussion of reasons for doing this and initial behavior may be found here. The specific post pointed to is a general overview, and the thread in which it appears contains much discussion and comment from active forum participants.

Installation instructions are here.

*********************

Okay, I failed to meet our moderator's (or akos') brevity standards, but this was just a trial text. I did not put a link in for installation instructions, as the one's on Trux's site are a bit skimpy, and to my taste the ones in the BOINC wiki are so conservative as to raise the pain threshold too high. If someone knows of a set of instructions written to the right level of detail (for example in the admirable style of Richard Haselgrove) please mention it.

To our moderator: Should you like this idea, please feel free to use as much or as little of my suggestion as you like, including pointing to different references.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.