Is there a problem converting pending WUs into completed ones?

> My impression is that it's
> not so much a bunch of slow machines causing the problem. Rather it's the
> statistic that there are now close to 150,000 registered machines but less
> than 65,000 have actually returned anything and been granted credit.

That is one factor certainly. At present also some people are affected by the effect described in the thread 'Why are they only assigning 1 of 4 sub-results of the WU?'. Where both effects hit the same user the delays will be even greater.

> It would be interesting to really understand why people go to the trouble of
> getting to the stage of registering their machine and downloading work but
> then don't do what seems to be the easy bit - let the box crunch for less than
> a day in total (even with a relatively slow box) and let the result be
> returned.

Actually it can be much more than a day. My 700MHz boxes take almost exactly a day to crunch a WU once I turn off the graphics, but if I left the graphics enabled they weould take much longer as those boxes do not have 3d graphics on the video card.

Part of the solution may well be to produce a much simpler (less pretty) screensaver that is the default option, and only cut in the 3d graphics when the user asks to show graphics. Alternatively it may be possible to make the client detect the presence/absence of a 3d graphics card and to only apply the 3d images in screensaver mode when the machine has a 3d card.

What I mean by a simple screensaver is a small text box that bouces round the screen and shows the progress of the current work unit, and the coords of the the point in the sky currently being processed, together with a note that advanced graphics will be shown when manually selected from the client. That way the machine is only wasting cycles looking pretty when there is someone nearby to appreciate the effort.

Given that most users (unlike me and thee) do not read the message boards and do not spend time tweaking the settings to get the performance/appearance balance right for their machines, I am not surprised that such users give up running the software before it ever checks in the first wu.

There are still a huge number of 700MHz machines out here, all doing useful work on their primary mission: obviously they are no good for modern games, but for those of us doing business apps (word processing, spreadsheets, etc) they are more than adequate and there is no real need to replace them.

Remember that also a lot of users will be trying to run E@H on a box that is only on 8*5 hours a week instead of 24*7: again no problem on a box with 3d on-card graphics or if they manually turn off the graphics; but getting even tighter if they leave the graphics running. Ironically then the machine does more useful crunching while the user is working than it does when they stop and let the screensaver kick in.

Having said all that, there will also be other reasons people drop out, people who meet a problem and can't be bothered to chase up the solution, as for every question on the 'getting started' board there are probably ten people with the same problem who never even asked and another twenty who never even looked at the board.

Then there are those who find that the initial downloads take too long on their metered dial-up connections (and who do not understand that the first download is always the worst on any DC project), those who never intended to stay but who download the software "to see what it looks like", and those indecisive types who just plain change their minds. This happens on all DC projects (or at least all I have been involved with). Typically after the first few times you submit work on any project you get about half way up the list of users.

One fix for those users would be a change to the scheduler so that it identifies newcomers (machines that have never yet returned a result) and for any given wu will only issue one result to such machines. With Three active machines and one newcomer there would still be a fair chance on the WU meeting its quorum.

> I have pending results going back to 2 Apr, my pending is now over 2300
> cobblestones, my RAC is plummeting. Or have I just been teamed up with someone
> running a 286 as part of the 'need 3 identical returned WUs' program? :)

Just ran across your post. Your question about being teamed with someone running a 286 caught my eye. (my machine is a 486, 600mhz). So, I did some research.

Guess what!! Of my last 11 results, 7 are shared with you. 2 of which I helped cause you a delay in credits by returning a client error - sorry- . 4 others we both finished successfully (one of these we are both still pending credit waiting on someone else), the other one is a WU our machines are working on currently.

The current WU should complete (for me) in 33 hours.

Another way to look at it - - I've caused a back log for you on two, and helped clear up 3 (or four depending on how you want to count the pending we're both waiting for).

Grace & Peace
WillTempo