Credits granted to large unit are not fair

Bruce Allen
Bruce Allen
Moderator
Joined: 15 Oct 04
Posts: 1,119
Credit: 172,127,663
RAC: 0

RE: The only thing I have

Message 62729 in response to message 62726

Quote:
The only thing I have proposed is that EAH and SAH would be my choices for the 'gold' standard for the reasons I mentioned before. It does not mean that those standards shouldn't be given a reality check review periodically, or that they are forever carved in stone.

You can find a cross-project credit comparison here. You will see that almost all projects are in good agreement with SAH. EAH was out of line. We're now fixing that.

(The credit comparison chart was created at my request by James Drews. It provides a solid basis for cross-project credit comparisons.)

Cheers,
Bruce

Director, Einstein@Home

Brian Silvers
Brian Silvers
Joined: 26 Aug 05
Posts: 772
Credit: 282,700
RAC: 0

RE: The only thing I have

Message 62730 in response to message 62726

Quote:

The only thing I have proposed is that EAH and SAH would be my choices for the 'gold' standard for the reasons I mentioned before. It does not mean that those standards shouldn't be given a reality check review periodically, or that they are forever carved in stone.

It does not appear as though your thoughts, which I agree with, are making any impact upon the decision-making.

Brian Silvers
Brian Silvers
Joined: 26 Aug 05
Posts: 772
Credit: 282,700
RAC: 0

@Alinator: I was in bed

@Alinator:

I was in bed and it suddenly dawned on me... There's another flaw in the logic being used to lower credits, specifically in regards to SAH "being in agreement" with other projects based on that chart...

The BOINC 4.x client underclaims

No matter what Ned says about the odds of pairing up with one of those clients, it would still be skewing the SAH project-wide claim/grant data downward.

I'm too tired to go into explaining all of it at the moment, but if you feel like it, go ahead...or if you, Bruce, feel like checking into it, that would be great too.

Annika
Annika
Joined: 8 Aug 06
Posts: 720
Credit: 494,410
RAC: 0

Wait... it underclaims? And

Wait... it underclaims? And here I was, thinking that SETI has been giving fixed credit for at least a year...

Winterknight
Winterknight
Joined: 4 Jun 05
Posts: 1,356
Credit: 358,925,031
RAC: 191,593

RE: Wait... it underclaims?

Message 62733 in response to message 62732

Quote:
Wait... it underclaims? And here I was, thinking that SETI has been giving fixed credit for at least a year...


Seti units has only been claiming the same credit/unit if the hosts where running a BOINC client that is capable of doing Fpops counting. Hosts that run older versions of BOINC still use the benchmark * time to claim, and these claims are usually lower.
And now that Seti has gone to 3 replications and quorum of two, where highest claim is discarded and granted is equal to lowest claim, there is some dissatifaction with the users running old versions of BOINC.

Winterknight
Winterknight
Joined: 4 Jun 05
Posts: 1,356
Credit: 358,925,031
RAC: 191,593

Regarding the cross project

Regarding the cross project credit comparison that Dr Bruce mentions, as Seti in the near future is moving to the multi-beam data, the comparison that Einstein should be using is with SetiBeta.
SetiBeta is where the multibeam data, with adjusted credits to be in line with Einstein S5R1, is being tested.

Cross reference to official mention of Seti going to Multi-beam is in Matt Lebofski - Quorum Change
Quote

Quote:
... As we ramp down the current data analysis to make way for the hot, fresh multibeam data, we are going to change the quorums for result validation very soon. .....

Andy

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2,142
Credit: 2,865,527,537
RAC: 1,436,907

Andy, nice to see you

Andy, nice to see you here.

I would be against using SETIBeta (or any Beta project) as a credit comparator. The initial credit allocations in a Beta project are inevitably a bit rough-and-ready. One of the final tests to be run in Beta, once the applications and server-side tools are running smoothly, is a credit calibration, so that the final release doesn't cause too much gnashing of teeth when it's deployed to the mass crunching community. (Another reason why it would have been nice, time pressures notwithstanding, if Einstein S5R2 could have spent more time in Beta).

IIRC, SETIBeta has a history of awarding higher credit than the main project - possibly as a deliberate inducement/compensation for the risk of running faulty WUs and getting no credit at all. When the last set of 'enhanced' apps were released for public use, the credit setting - the FLOPs multiplier - was reduced for parity. Caused a bit of angst when some crunchers started using the Beta application in the main project.

Winterknight
Winterknight
Joined: 4 Jun 05
Posts: 1,356
Credit: 358,925,031
RAC: 191,593

Hi Richard, Have to move

Hi Richard,
Have to move arround sometimes.
I can see where you are coming from, and for the last hour or so have been trying to find a cross reference post about SetiBeta adjustment of credits to be in line with Einstein. I know it is there just cannot find it.

I also don't think Seti is the project for Einstein to be comparing with, unless they can compare credits/time using standard application only. I'm not sure the refered table does that. And adding the fact that Seti is going multibeam, very probably with the credit calculation used in Beta. Then comparing with recent/present Seti is only going to add confusion and possible anger to the pot again.

The other question that springs to mind, is, over what time period is the cross project credit comparison taken? long - since projects started, medium - last 6 or 12 months, or short - last month or week.
If it is medium(ish) or short then SetiBeta is the right project to compare to, as most users crunch with the standard app, or similar (e.g. Joe's).
Comparing to Seti overall brings in an unknown mix of standard/optimised apps.

Your assumption that trial projects give inflated credits I don't believe is true look at SetiBeta host 351 result from early in trial resultid=598190 under 2cr/hr on Pent M.

Andy

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2,142
Credit: 2,865,527,537
RAC: 1,436,907

RE: Hi Richard, Have to

Message 62737 in response to message 62736

Quote:
Hi Richard,
Have to move arround sometimes.
I can see where you are coming from, and for the last hour or so have been trying to find a cross reference post about SetiBeta adjustment of credits to be in line with Einstein. I know it is there just cannot find it.


Would this one be a good starting point?

Edit - scroll down to the bottom and have a look at the OP for that thread too.

Winterknight
Winterknight
Joined: 4 Jun 05
Posts: 1,356
Credit: 358,925,031
RAC: 191,593

RE: RE: Hi Richard, Have

Message 62738 in response to message 62737

Quote:
Quote:
Hi Richard,
Have to move arround sometimes.
I can see where you are coming from, and for the last hour or so have been trying to find a cross reference post about SetiBeta adjustment of credits to be in line with Einstein. I know it is there just cannot find it.

Would this one be a good starting point?

Edit - scroll down to the bottom and have a look at the OP for that thread too.


Maybe it gives me a date to start from, there is a post with the actual credit adjustment numbers in it. I think it went from 8.x to 9.n. And I'm pretty sure it was on the SetiBeta site.

Comment viewing options

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