The only project running on the World Community Grid is very CPU demanding. But if you look over at http://www.grid.org , you can specify their Cancer Research Ligand Project, which uses smaller Work Units. The WCG was modelled after grid.org, so you already know how to install and run their Agent.

Also, probably because they have a lot of older computers, they have stretched the time limits for Rosetta to 3 weeks CPU time and 4 weeks wall clock time. [Added by Edit: grid.org is currently running 2 projects. One of them is Rosetta, which is run in conjunction with the World Community Grid, but uses its own requirements. A CPU must have 256 MB RAM to run Rosetta, but is allowed to take 3 weeks run time to complete a unit. WCG only requires 128 MB but insists that the Work Unit be completed in 2 weeks of run time.] This means that potentially a batch of Work Units from the ISB could take longer for them to process because of a few laggards. On the other hand, they have more computers in their grid. So everything is a trade off.

You might want to try out a few Ligand units on your 600 Mhz and see how long they take. Hopefully we will eventually have a mix of projects of various sizes here on the World Community Grid.

Lawrence

A little off topic here ... sorry ... but your dilemma begs the question of will the grid server reassign the same WU to yet another machine that cannot finish in 2 weeks?

If the grid server does NOT take into account the agent's "Device Overall Performance" score then those WU's that are long running will bubble-up to the top over time, and so the longer the project goes the longer running will be the WU's asssigned. Towards the end there could be some very long running WU's that will have been unfortunately reassigned over and over and will be tough to finish. It would also be discouraging to users since they may feel like they are wasting their time just when the project is nearing completion, and they are unable to complete most of the WU's they are being assigned, and no way to request a newer WU.

But am I wrong? .... when a WU is reassigned is the device's score taken into consideration?

Is there a way to request a newer WU when the user see's the current WU is a lost cause?

Is there a place to go to see when a WU was started, so someone can see how long it's been since (s)he started? I currently have a WU that's accumulated 25+ hours of run-time, but I don't remember what date I started. Was it five days ago ... maybe 6? If I knew what date then I'd know to let my machine work 24x7 so as to not waste the effort as the 2 week limit approached.

John

There are currently two different timeout values:
* 2 weeks of Run time
* 3 weeks of wall clock time.

These runtime values are project specific. The next project may use different values.

This means it will abort the workunit if you've actually spent 2 weeks crunching it (336 hours of Run time on the agent). It will also abort if you've had the workunit for 3 weeks regardless of how much time your computer has spent working on it.

Other than making sure the device meets the minimum requirements, the device capabilities aren't taken into account by the UD Server Software.

Finally, there is no way of knowing ahead of time which workunits will take a long time to complete and which will be quick. They're non-deterministic. That's why this project is on the grid...to do hard computational processing.

Many thanks to lawrencehardin, Alther and Graham for their timely responses. I will print this off and save for future reference. :)

Now ... I don't want to rain on anyone's parade ... but the info you guys have given me may be a little unpleasant to some of the WCG users. I currently am using a 3 Ghz processor on a WU that started 4+ days, has accumulated 28+ hours run-time, and is only 34% done. If this trend holds up it will take me about 84 hours to finish it. I don't anticipate any problem finishing this WU myself, and I don't mind, but someone with a 1.5 Ghz computer would take about 168 hours. Over a 3 week period this would require 8 hours a day of run-time. Many of the WCG users are casual users and only on their computer for an hour or two a day. I'm concerned the project as a whole could get a black-eye if these users spread the word the project is bad/useless/waste-of-time/etc. In Alther's post he said "making sure the device meets the minimum requirements", but the requirements page merely mentions a Pentium CPU (did I miss something?). To avoid bad word of mouth, and wasted WU allocations, consider adding a requirement that the user have a Pentium of a certain caliber, and maybe even a willingness to devote a certain number of hours per day to the project. Let them know what they are getting into.

This is a wonderful project, and has many implications for mankind, but we shouldn't expect grandma, who logs on once a day to check for e-mail from her grandkids, to understand the needs of the system and have to explain to her why she's wasting her time. God forbid my 82 year old father ever hears about this project and starts asking you guys questions. :). You think I'm a nuisance now ... wait till he get's his shorts in a wad. :)

The point is guys and dolls ... a grid project such as this is for the masses, not just the technically gifted ... and they're needs are much more so than we are used to.

Thanks for everyone's time.
John