I think it has been said before, but what I would like to see is the number of "active" devices. Maybe even active devices per project. Active can be defined as returning a result within the last time-out period (say a week/ 7 days). There is a webpage out there that displays stats based on yesterdays numbers, and taking the time output and saying how many devices we have.
http://vspx27.stanford.edu/DCcomparison.html

Also how about the average processor score for all of the active devices. If you have that information.

Can someone verify, I think I saw at least a 1,000 device increase since the linux client release. Tracking active devices would tell us how many of those may have been running the windows client and are glad they can switch over to linux and others have boxes on the side that were doing nothing for WCG, waiting for the linux release.

Also, I havn't been seeing people do those calculations to tell how powerful WCG in terms such as Flops, etc. We know that we are more powerful than super computers, but how powerful? This information can be used in future press releases. They love facts. Give them as many as you can.

I personally don't like the points system. I see one of my systems, that I know is slower have more points, just because it has more time. The other, which is faster and done a lot more work, still has less points. The only way that I can think of for this to be correct is if the slower computer did harder work units than the fast computer. I hope someone understands what I am trying to say here.

With some older systems, to me, it is better to save your electric bill and upgrade.

Hi Sterling2p,

you're right about those hard facts. Almost every project has them.

About the point system,
I found out that there's an easy way to 'cheat'. It's not exactly cheating but i'll explain.
When you run an instance of the WCG app and run another DC project on top of it that runs with a slightly higher priority it will consume all idle CPU power but the wall clock time keeps counting. Then after a few days or so, you let the WCG workunit complete and let the benchmark have a nice result.
Points = Wallclocktime * Benchmark
So while you actualy crunched about 3 hours (just to say something) you claim points for 4 days. In the meantime you can get credits for the other DC project as well. (I can think of some people who are stats addict and do these stuff , not looking at myself ofcourse. I'm in for the science but the stats are helpful to keep everything interesting)

Hope i didn't make big mistakes in here

Oops,

I finally got around to reading the papers about BOINC at http://boinc.berkeley.edu/papers.php David Anderson seems to have developed a second-generation points system that looks pretty good. The processor benchmark is run several times during a long work unit, which should do a much better job of tracking CPU utilization. But the big improvement is that points are not awarded when the result is returned. Instead, the server waits until the validator daemon has enough copies returned to declare them valid. Then the scores of the valid results are computed and, at the option of the project, either the average score or the lowest score is awarded to everyone returning a valid result. So everybody gets the same number of points for that work unit. This is fair and makes cheating very difficult.

I should have read the manual before pontificating!!
mycrofth

It would not be a problem to eliminate devices that are more than, say, a rolling 3 months old from the current date, that have NEVER returned a result.

That keeps the new devices around that may be having some sort of temporary problem.

If you eliminate old devices that HAVE returned results, you ruin the history.