Upon checking the workunits I currently run, I found two different tasks which have exactly the same name:

658255737
658255719
name: rb_05_04_46751_91195_ab_stage0_t000___robetta_IGNORE_THE_REST_09_09_163401_92_0

For both tasks I have currently one workunit in process on the same computer.

1.) Is this normal with Rosetta@home?
2.) Does it mean I'm doing the same calculation twice here?

Thanks in advance for any insight.

PS: The reason for checking all my workunits currently is that I've lost several days of work at least on one computer, since the server flagged all workunits in process as 'client detached' and I got no notice whatsoever on the client side. That's most unfortunate, especially since the credit is lost for the Pentathlon. I already canceled all work on that particular computer (example).

Hi Nightfall,

Sorry about the lost work. I'm not sure what happened here I'll ask around what may have happened.

But as far as the same name goes. Even though the names and input is the same, the result and calculation performed is different.

An analogy: Lets say you want to find the tallest mountain, but you do not know where it is. You send a minion (from Despicable Me) to find it. The minion runs around and finds a hill, getting excited as it climbs higher and higher eventually reaching the peak!! The minion is happy and it's job is complete. But the problem is, the minion only found a local hill, and this is not a global peak =[

Instead of sending one minion, what we do is send off thousands of minions distributed randomly across the globe, with the hope that some of them are close enough to the tallest mountain to find it. Sometimes if we know some prior information (as in "what region might have the tallest mountain") we restrict the sampling in this area.

The two jobs with the same names, are the two minions released from two different starting points but the same landscape. The work of each minion is important to insure that we found the global peak! =P

Thanks, krypton, for the analogy. Nice one. :) Knowing a little bit about the science behind Rosetta I was hoping for something like that. I just wasn't expecting exactly similar task names even if the same dice is cast twice on the same table.

Both workunits were finished successful, btw, and the Pentathlon discipline Rosetta ended two hours ago.

Thanks for taking the time.

Many BOINC projects involve complicated calculations that are easily verified, e.g factoring large numbers. Just multiply the (alleged) factors and the answer is either right or wrong. That makes it easy to toss bogus results that an evil participant might return.

My understanding of Rosetta is that everyone takes a random walk through the space of possible answers while trying to get closer to the expected answer. Since the models aren't perfect and the calculations involve rounding errors, there won't be a "correct" answer that is exactly right. Although it is relatively easy to measure how close a result is to the expected values, there is no practical way to determine the closest possible answer that the model might generate. Thus the same task is farmed out multiple times and the closest result is accepted.