I'd be in favour of anything that improves overall turnaround times but, unfortunately, I'm not sure that just [say] doubling the number of simultaneously active WUs will show a doubling of completion rates :-)

It would be nice to know exactly what the Project Statistics page counts as a returned result. Unless it's everything that's sent back (errors and all) the proportion of excess results against WUs processed seems to run at a steady 10% or so (after allowing for Extremes probably returning three results anyway). Given that not every missed deadline task will become a late returner that validates, there is probably an even higher proportion of WUs with tasks going to systems that are going to miss the deadline.[*1]

Those tasks could, instead, be going to users that have no problem returning work on time. That's why some folks are in favour of doing one or both of

• capping the number of concurrent ARP1 WUs any host can have;
  
• reducing deadlines (perhaps with a grace day to compensate in part).

Note that reducing the deadlines wouldn't actually stop "small" users with older hardware from running tasks in a [relatively] timely fashion (provided they don't set up with a 5+ day buffer!)...

It's all a matter of opinion (I'm pro-capping and grace days [with shorter deadline] myself); as it happens, I doubt anything will change in those respects :-)

Cheers - Al.

*1 Missed deadline responses would be a useful statistic if based on the actual databases! Yet another time-consuming wish-list item :-)

Thanks for giving us the numbers Adri. It is nice to see the ARP WUs flowing better. yesterday the reference number in my thread was close to 400 but now it is almost 500, so they are still slowly ramping up and sending more out.

I have 5 of my 10 CPUs reserved for ARP, but I'm lucky to get one now and then. It is a race to see if I'll get my 2 year badge in ARP or MCM first. I've handicapped MCM to only 1 CPU, but it still might win.

Unixchick

What are you doing on your other 4 CPUs? Are they different machines or do you mean threads?

It is difficult to get ARP if a machine is only working on ARP because of the long spacing between uploads. Mixing ARP & MCM on a machine would result in more frequent uploads and get you more ARP downloads.

Mike

Unixchick

Just to clarify, by cores do you mean threads?

So far as your other uses are concerned, most would have intermittent use of a thread and WCG allows access for those tasks so you can bump up your usage.

For example, if you are typing up a spreadsheet, a thread is not actually used until you press enter. BOINC would release the thread momentarily.

As for my daily reports, I was thinking that I might drop them once there are no longer any accelerated units moving. I would continue my Sunday Reports as they serve a different purpose - the forecast, for example.

Mike

When you say "there are now 13,511 held up in generation 143" what exactly do you mean?