Sad thought!!

why not just tell the wingman we don't want their help

by my calcs he missed 6 hr/60% cutoff by about 30 min, you made it by about 30 min, pretty good matching. because of this you got almost 50% more time to finish.

As for resends/repair jobs are not sent to "known high performers" they are sent to reliable performers. my 600Mhz box does get repair jobs, if it had landed there, you would have got approx 10% complete after 6 hrs, by your system we throw that out also and resend again until results match your desired standard, appearing to me as possiably much more waste.

as for telling crunchers, "your results are not as fast as your wingman, so we are throwing it out and redoing it". How many do you want to get rid of? I for one would just quit "Wasting" my cpu cycles and find a better place for them.

In my nearly forty years of IT - from programming, analysis and system design - I have noticed a consistency.

I'm brought into a failing project and the managers all run away, pointing at me. I get no co-operation and devise new techniques to solve the problems.

When I've got the problem licked, I can't fight the managers off with a stick. They'll criticise my techniques as being "non-standard" or "not approved."

At the very best outcome, the system is forced to run at only a small portion of its potential. Sometimes, it fails completely because the managers want to 'rationalise' what I did - without understanding the fundamentals.

The more the managers ignore me or pooh-pooh my decisions and techniques; the more the responses are political rather than technical, the greater the probability that the latter outcome will be the case.

The commonality is this : the project is inevitably managed using "politicians' delight" arguments - those that are simple enough for a politician to follow, theoretically solve the problem and will survive a cursory examination but which will fail when implemented, needing more "big-noting" executive decisions from the managers.

As I have said before, the solution lies in splitting the longer returned unit into separate smaller returns and transmitting ONE "make-up" unit rather than two - and we're aware from the later-generations numbering on this project that units can be created of any arbitrary size. Sadly, I had no technical response to that proposal.

I am sure that if what I've suggested was implemented, wastage would be eliminated. It would obviate the need for matching and could be applied to ALL processing.

The Crunch-till-you-drop scenario would apply only to a minority of crunchers who CHOOSE to adopt that option. It would not address the wastage as far as the other crunchers are concerned. It would mean that the report deadline would have to be extended for CTYD adopters and does nothing for the other time-waster - "too late" units where new tasks are despatched while crunching is still in-progress(*)

CTYD would take as much programming effort as would implementing what I have proposed in my estimation (and after 40 years, I believe I have some vague inkling of what I'm claiming.) It would also only suit a minority, albeit a vocal minority, rather than being a solution for all.

But after forty years, I'm also quite used to being shouted down, criticised, pooh-poohed and talking to brick walls. It no longer worries me - I can only do what I am allowed to do and I don't want to get involved in the politics. Not my area. I'm a technician and only understand technical arguments.

Sad to see that you propose to leave HCMD2, martin64. Personally, I'm on a fortnight's excursion to another project; back soon.

(*) yes, I know that this would require BOINC changes to report the current progress of a unit and is thus outside of current project-control.

I started tracking the valid results for this projects as when this topic got revived. On the first 21 (W7 and Linux), there was 1 that had a combo of 6 hours and wingman a longer time. After 47 valid I've now 2 with this condition and 1 that met as 6 hours with 6 hours, with relative close credit award. The sample size is relatively small, but project wide I would not guesstimate it to be far off and thus the by me 'perceived' waste being relatively small.

I'll keep tracking on a daily basis so as to maintain some feel for what my devices produce.

Now within the skill that BOINC actually has, on chopping HCMD2 work units up to the point where every unit can be completed no matter what clients meet, it can only mean it has to be done per the lowest common denominator (so I understand). That could mean an exponential growth of work units to distribute which in turn might impact the overall grid performance. As it stands we're on around 700,000 daily average and no idea how confident WCG is pumping this up to a million and more with several projects coming into production and HFCC returning in a little while. The schedulers/feeders/assimilators will be very busy me thinketh.

The "85%" I only found in relation to [from the outside showing] credit disparity. See this http://www.worldcommunitygrid.org/forums/wcg/...ead,27770_offset,0#256089 and the follow up post.