On Classic, I periodically would collect all of my user stats and put them into an Excel sheet set up to run little calculations and stuff to track my progress for different segments of time and graph how much work I did and how it varied and stuff. Basically for fun (and gave me something to do when I was bored), but I used a variety of computers in different amounts and I just wanted to see their effect.

I hadn't really planned on doing that so much in BOINC since some of the stats I want aren't readily available in one place and the credit is per computer, but I finally decided to go ahead and start a new one up. In setting up the file though, I found what appears to be a discrepancy in the RAC.

The 4 computers currently tracked by BOINC (really three but 2 can't be combined even though they are the same computer) report RACs of:

142.5
63.24
58.36
54.36

That adds up to 318.46. Yet my account reports me as having an RAC of 278.64.

RAC is calculated based on average daily credit with a 2 week half life (or something along those lines). I don't really know anymore about how it is calculated, but I don't see how that would cause each computer's RAC to add up to something different than the total RAC. I could understand a few hundreths maybe for rounding, but a difference of 39.82? Am I missing something here or is this a bug somewhere in the system?

>
> RAC is calculated based.. Am I missing something here or
> is this a bug somewhere in the system?

an endless story. slower cpus reports a higher RAC than faster.
there is so much "unknown", you will probaly never see, like on seti 1,
what are your system "realy able" to do, the values at berkeley will never
reflect your work.

u can stopp crunching WU's and still RAC can increase for a time, until all credits are spoken.

==> If you use a "management" software, par example BoincView, you can inport the log file(csv) an begin the exel work.


Due the "design" of granded credits and the way they just can't reflect our "efforts", I changed my focus:

From the source of the client, I took the cobblestone formula and learned,
how it works.

taking the values from the boinc benchmarks, now I'm calculating the "estimate", how many credits one PC can do in 1 day.

ric

It's not the amount of credit I have. It's that the amount of credit reported on the main account page for recent average is NOT the total of all of the RAC reported for each individual system. Depending how they calculate it, I could understand a few hundreths off, but this is off by a significant amount and either one or more of the numbers are wrong, or I am just missing something.

Chris

> Due the "design" of granded credits and the way they just can't reflect our
> "efforts", I changed my focus:
>
> From the source of the client, I took the cobblestone formula and learned,
> how it works.
>
> taking the values from the boinc benchmarks, now I'm calculating the
> "estimate", how many credits one PC can do in 1 day.
>
> ric
>
>

I too have seen this many times. It was pointed out to me by someone on BOINC-IRC, and has baffled me ever since. There is probably a good reason...

My theory is your USER RAC is actually updated when a work unit is uploaded. "Updated" how though is the question. Is it simply added? Who knows. Because the host and user xml files are updated at different times, maybe there is a factor in there somehwere.

*shrug*

(On as side note Richimiester, I got your email but can not reply cause the email you listed BOUNCED... if you want a reply just send me another email with your correct email addressA)

---


SETI Synergy Webpage

BOINC World/Team/Indiv Stats

RAC is like a speedometer, it shows you how fast you are getting credit. Your RAC still goes up after you stop crunching because that is how the granted credit system works. Sometimes you will be the last person to return a unit and as soon as the computer recognizes that you will get credit, alot of the time you WILL NOT be the last person and you will have to wait for someone else to return the unit before credit is granted.
Clear as mud now?

Why make things simple when we can make them complicated...

Since this is good and confused I'll through my 2 cents in and make it more confused.

RAC is calculated when something changes ie. a workunit is returned/granted. It is also recalculated on a schedule (about once a week I think but with the server problems all bets are off). So the total will not match the total of all the individual machines most of the time since the total gets updated every time any machine gets updated but an individual machine only gets updated when that machine returns something.

To fruther complicate matters I don't think there is any kind of cross checking between the two. When credit is granted the total and the individual are adjusted seperately. Since they use the same fomula they will stay somewhat close but since they are computed indipendately they will never be exactly the same.

I hope everyone is as confused about this as I am now.


John Keck
BOINCing since 2002/12/08

The way you described that, it still seems like they should be the same (or within a few hundreths). But it seems you may be on to something, especially with the server problems. Since RAC is calculated for each machine though, it would simplify things greatly and make them essentially failsafe to just have total RAC be calculated as the sum of each computer. That way they would change together and any percieved difference (if any) would come from rounding individual totals for display when necessary. Then it would at least make some kind of sense.

On well, at least you, Zain, and and maybe Little Sun understood what I was getting at and are as confused as I am about it.


> Since this is good and confused I'll through my 2 cents in and make it more
> confused.
>
> RAC is calculated when something changes ie. a workunit is returned/granted.
> It is also recalculated on a schedule (about once a week I think but with the
> server problems all bets are off). So the total will not match the total of
> all the individual machines most of the time since the total gets updated
> every time any machine gets updated but an individual machine only gets
> updated when that machine returns something.
>
> To fruther complicate matters I don't think there is any kind of cross
> checking between the two. When credit is granted the total and the individual
> are adjusted seperately. Since they use the same fomula they will stay
> somewhat close but since they are computed indipendately they will never be
> exactly the same.
>
> I hope everyone is as confused about this as I am now.
>
>
John Keck
> BOINCing since 2002/12/08
>

> The way you described that, it still seems like they should be the same (or
> within a few hundreths). But it seems you may be on to something, especially
> with the server problems. Since RAC is calculated for each machine though, it
> would simplify things greatly and make them essentially failsafe to just have
> total RAC be calculated as the sum of each computer. That way they would
> change together and any percieved difference (if any) would come from rounding
> individual totals for display when necessary. Then it would at least make
> some kind of sense.
>
> On well, at least you, Zain, and and maybe Little Sun understood what I was
> getting at and are as confused as I am about it.
>
The value of the RAC depends upon the time that it was calculated. If machine A had a RAC or 50 last week, and has not been recalculated since, its RAC will show as 50. If machine B has just calculated its RAC and it is 50, it will also show as 50. The total RAC would show something like 75, as the current RAC for machine A is actually about 25 rather than the 50 shown.