Quorum question

[RAMMIE]

Fast vs slow machine.
What is a better approach for max points for each?
1=Be a member of quorum and set the bar.
2=Ride along and sop up whatever points are given.

[ShootStraight]

Better approach? I'm not sure I understand what youre getting at.

Could you explain?

-SS

[sierra_bound]

Ideally you want to be the quorum member with the lowest claimed credit and go along for the ride. Every now and then, I'll have a WU with a claimed credit of, say, 70. But I'll end up with a granted credit of 90-100, thanks to the other quorum members.

[ShootStraight]

Ideally you want to be the quorum member with the lowest claimed credit and go along for the ride. Every now and then, I'll have a WU with a claimed credit of, say, 70. But I'll end up with a granted credit of 90-100, thanks to the other quorum members.

First, S_B, Can you explain the rationale behind this statement? Maybe I'm being obtuse, but that makes absolutely no sense to me - not saying youre wrong, I just dont understand.

Second, I'm not seeing how there is an "approach". That to me implys that there is a cognitive choice involved which I'm not seeing - What am I missing here?

-SS

[sierra_bound]

Check the screenshot in the link. It explains what I'm talking about.

http://www.xtremesystems.org/forums/showpost.php?p=2090171&postcount=1

In this particular case I had the lowest claimed credit (48.4), but got granted credit of 56.4, thanks to the claimed credit results of the other quorum members.

[ShootStraight]

OK, I get what youre saying. You got an extra 12 points whereas they lost 9 (at worst) over claimed credits.

However, it would be even better if you had claimed closer to the high of 65 - this is what confused me as the points granted would be higher.

-SS

[sierra_bound]

It's all a matter of luck. You just have to be in the right quorum. In the majority of cases, my granted credit is usually higher than my claimed credit.

[[XC] riptide]

I have postulated before that with the current current Quoram members I have found that a Dual Xeon (Netburst) at 2.8-3.2Ghz will generally be granted very close to what it claims everytime. I suppose this would also apply to P4 2.8-3.2Ghz. Would that tell me that currently the VAST majority of WCG members are on this scale of machine?

[meshmesh]

The quotum system works on average claim. So there is no bar setting.

Claim low and you drag the quorum down with you, hurting yourself and the other two quorum members.

Claim higher than the other two, and every one gets a bigger granted credit. On the other hand, too high and you get discarded from the quorum calculation and the other two machines will set the quotum. And it doesn't take much to be eliminated.

The best strategy is to ensure that your machine claims close to the average of the whole population in order to be included in every quorum.

Unfortunately, although it is easy to ensure that the machine claims correctly relevant to other crunchers running the same family of processors and same OS and client, it is not possible to ensure all the time that it is near the average for the whole mix. But the error is generally small and evens out.

Machines based on C2D (and it's derivative) occasionally claim very low Iops. Although the quorum mechanism will exclude the machine if it is too low, it is still not something that should be left uncorrected. There is always a remote possibility of getting another bad-benched machine in the same quorum. In this case, the awarded credit will be really low. With the new price drop the percentage of C2Ds will increase in the population. Hopefully by then, everyone would be able to move to a new stable Boinc client that solved this problem.

Since Intel has about 90% of the market share, and since most machines are over a year old, it follows that most machines out there are based on P4s. Large boinc projects would reflect this population composition. Come 2008, things may start to change.

[sierra_bound]

Claim low and you drag the quorum down with you, hurting yourself and the other two quorum members.
This is not always the case. I've had a few WU's with a claimed credit of 70 and ended with over 100 granted credit, thanks to the other quorum members. 70 is not really a low number. As I mentioned earlier, this is purely a matter of luck and it doesn't happen very often.

To be quite honest, I don't pay a lot of attention to this. It's just something I've noticed from time to time. I'm more interested in WU's submitted and validated because this benefits both the science and me.:)

[Movieman]

another approach:
The hell with how the quorum effects you.
Sometimes you do better than you should, sometimes worse.
Hopefully in the long run it's a wash but even if it isn't, it's beyond our control.
The faster machine does more WU a day and gets more points..period!
Crush them with mhz's!:D

[[XC] serlv]

another approach:
The hell with how the quorum effects you.
Sometimes you do better than you should, sometimes worse.
Hopefully in the long run it's a wash but even if it isn't, it's beyond our control.
The faster machine does more WU a day and gets more points..period!
Crush them with mhz's!:D


That is my approach as well.

[meshmesh]

another approach:
The hell with how the quorum effects you.
Sometimes you do better than you should, sometimes worse.
That is a very good approach. However, I also keep in mind that being a member of a quorum of three, I owe a duty of care to the other two members. I may not care about the points, but other project participants do. It is for the benefit of the project to retain as many members as possible.

I don't need to spend too much time on this. But whenever I check my results and notice that my machine is consistantly under claiming by a significant margin, it warrants an investigation. That's all.