Today we just passed 44,400 years of total processing of FAAH. Experiments 1a and 1b ended at 14,400 and Experiments 2 through 5 started. Since then 30,000 years of processing has occurred and we are very close to the end of experiment 5.

You'd have to go thru the whole thread to get the history and how he got there. On the original scope and extrapolating from a different representation of phase 2 on the project website, in Work Unit terms we're somewhere over 95% complete.

Later in phase 2, the project was split up in 5 experiments (2-6). For all we know, the scientist could have not updated the progress charts for experiment 5 and 6 and we're already halfway the final experiment ?!?

.... Anyway, the newsletter of March 07 advises that FA@H will continue thru Dec.2008 which is of course an estimate. That work is going to be running as lets call it provisionally Phase 3. When exactly phase 2 ends and phase 3 kicks in... don't know!

lawrence' *speculation* is accurate. There is one person, who's in charge of the work planning (see the graphs a few posts up showing the history)! There's only a limited resource on programmers that can take a projects science and put it into the BOINC or UD agent frame. That amongst to a large extend co-dictates what is being planked and what not.

As for user preferences the rule:

First give work full time to those selecting project X as sole choice **
Second give work on a shared bases to those with a limited project selection
Third, put remainder in pool and share it out based on available volumes

CPU time and quantity are factors here that co decide where goes what.

The FAAH slowdown is largely driven by the need to accelerate HCMD as it will go on a 4+ month break in preparation of phase 2 (Sept.07 target).... it bought time for the next phase of FAAH to complete thorough testing before go-life.

** Make sure that users understand that work can dry up like happened on HDC regularly, thus the 'back up project option' is being selected.

65-70% of the crunching is on UD Agent still, thus pushing HCMD on that agent will impact FAAH overall substantially.... but as said the timings are thought to be set so that when the next phase is ready for launch (with which the scientists and programmers have their hands full), they can focus on the final results of the last experiment. What's the point of rushing the data back to Olson when nobody can look at them whilst HCMD e.g. is pressing!

It's all theory and conjecture, but this is how i would manage the pool.

One BOINC machine of mine on a mixed profile of 3 out of 4 available projects, is only getting FAAH momentarily in the last 48 hours or so
(does 7-9 a day), which is more or less the old 'every 8 FAAH, 1 of the rest' mix.

So far, WCG tries to keep projects that are phased, to continue seamlessly or a little overlapping as in one winding down, the next firing up. Actually in the below example, HPF1 finished about mid June, but there was overlap where some WU's were send out to complete the batches for HPF1.


Why the conclusion that phase 2 of FAAH is futile? As said, the scientists have hands full in prepping the next whilst the current is running towards completion.

Anyway, I let lawrence answer anything more on the matter.