I opened a discussion on the CPDN forum (http://climateapps2.oucs.ox.ac.uk/cpdnboinc/forum_thread.php?id=3965) which developed in something which is better discussed here.

Debt management is maybe the most advanced topic handled by the scheduling process. It has to achieve the best scheduling in slow and fast computers, with one to more than a dozen projects attached, and workunits which request from ten minutes to one year... Balancing workloads, keeping deadlines, respecting resource share defined by users...

Such a critical aspect is however "not visible", in that whether I am devolping an enormous Long Term Debt with CDPN or not is not shown.

You might find interesting what has been said in the above-referenced discussion, and maybe contribute your own thoughts...

---

John McLeod VII is the man you need, he's the volunteer developer who suggested it and was encouraged to program it, he will be the first to admit it is not perfect. But it's a lot better than what we had before.

I've read the thread on CPDN and I'm pretty sure that only the short term debt (STD) is adjusted when a project is suspended, the LTD still carries on.

In the long term it doesn't really matter if your computer has decided to process only one project at the moment as it will build up a large LTD and then not download ant more work for that project until that debt is worked off.

It is almost certain that running that many projects your computer is going to running one project at a time because of the scheduler rules. If there is work available for all nine projects and you have work from all of them, on average they are only going to get 2h:40m each/day (if resouce shares are equal) and at that rate your climate unit will need about another 60 weeks to complete, (if the figures on original thread are correct). The computer will at some stage have to crunch that unit exclusively unless you detach from or set other projects to 'no new work'.

To run that many projects with varying crunch times and deadline lengths, I would suggest learning the scheduler rules, several sharp pencils and a new pad of paper to work out what the resouce shares and 'connect to network' options are to be set at to avoid not crunching one unit exclusively.
And even then its all going to be turned upside down because of 'no work available', server or comms downtime, plus your own brain fades and finger troubles that we all suffer from.

Andy

---

Let's make something clearer... ;op

1) If John McLeod - or anyone else - is the man behind the current attempt at "clever scheduling", then he DOES deserves my esteem! I think that component is extremely challenging. Maybe, it's the most important single component in whole BOINC project; IMHO, it is the most ambitious work in the current BOINC framework...

2) In no way I wanted to criticise work done by others. After all, they DID something, while all my effort was to launch the application and, at most, run some configuration...

3) All my point could be synthesised in "gise us some (blinkin') light!". I'm not saying something is wrong (nobody is selling me a finished product, we are all running an "in development" project. I am saying that in my case, implemented behaviour is not optimised. Since
- I'd like to give some more feedback (and, also, to understand a bit more what's going on... ;op )
- I already have tuned allocation of resources to the different projects (for your reference: Rosetta, Simap, BURP and Predictor are at 6,25%; LHC, Seti, SetiBeta and uFluids are at 12,50%; and CPDN alone is at 25%)
- I have not the time or the will to go through documentation, forums and so on in order to understand the whole thing more then I currently do.
- Debt (both long and short) data is known to the client
Since al this, I said, *making debt data visible* seems to me both easy to afford and efficient.

After all, I'd simply like BOINC to say something like "Hey, you! I am not downloading anything else because U mistreated this project and I, The Client, am now making justice!"... ;op

Thanks for your work, guys!

PS
Also adding two columns to the "projects" tab, one for LTD and one for STD, would do... ;oppp

---

Also adding two columns to the "projects" tab, one for LTD and one for STD, would do... ;oppp

Have a look at BoincView. It will show you a lot more of the internal data the BOINC-client is working with.

Norbert

---