Vijay explains why.......

[Hugo 59]

This was posted by 1g_Raider at aoa thought you guys might want to know.

the work units are taking longer than usual.

------------------------------------------------------------

WUs are taking longer than we originally expected. This is not
uncommon in protein folding. When the protein gets collapsed, many amino acids are interacting with each other and the calculation slows down. It's hard for us to predict when this will happen (that's in part why we're running the simulations).

Everyone is in the same boat though, so you're not losing credit
compared to anyone else.

Vijay

This was posted by 1g_Raider at aoa thought you guys might want to know.

[JetMech]

Thanks for the heads up Hugo 59.

[1g_raider]

Here's a little update on this subject.

Vijay did a little pole at the Yahoo e-group and made a decision on work unit length. Here is his conclusion to the pole.

---------------------------------------------------------------

Hi,

Thanks to everyone for responding. The overwhelming response was for shorter WUs. The biggest problem with short WUs for us is the added burden on our servers. Right now, we could try to make it a little shorter for future WUs. We can't change in midstream, but I'll plan to try to make the WU's around 1/4 the current size. I think that's a reasonable compromise. Also, keep in mind that shorter WUs are bad for people with modems and I have to balance everyone's needs. Science-wise, we have some flexibility here (multiple short = few long).

So, we'll try it for future WUs, which should be coming on line next week or so. We're running villin as a validation of the system and to finish off that data set. We should be moving on relatively soon (1-2 weeks).

Vijay

----------------------------------------
Now all we need is the ability to cache a couple of w/u's and we will be cooking. I hate it when you have been running a w/u for 3hrs. and it wont checkpoint. Now maybe it will be a little more tolerable when the core F's up.

Just a heads up for you guys.
:)

Oh, Mike - feel free to repost anything I have posted over there. I just want to help where ever I can. ;)

[lennytiger]

souns interesting when will the new proteins be here what will they be called??

[1g_raider]

Heres the bad news:
He's takinig about messing around and cutting the points rewarded for the shorter w/u's.
-------------------------------------------------
That's the right idea. More specifically, right now, villins get 2
pts. If we send WUs with 1/4 the iterations, they'll get 0.5 pts/WU.

Vijay
-----------------------------------------------------

I don't think this idea is going to sit well with the masses. I hope he decides to rethink this.

[1g_raider]

Originally posted by lennytiger
souns interesting when will the new proteins be here what will they be called??

No info yet on both of your questions.

[Kendan]

I don't see why this should make anybody mad as long as he keeps the point system the same. He benches them on celeron 400's if they take 24 hours you get 1 point,48 hours 2points,12 hours 1/2 point. But he is in charge so it might be changed. I don't really care but shorter WU's would be nice because I just lost a 3/4 finished villin because I had to reboot my rig without closing core_65. Fold on!!!

[Ploaf]

I personally think that they should be at least one point and go up from there, but it doesn't matter either way. The current villins could be worth 4 points then. The main benefit to breaking this down more is that the project may advance more quickly. I was losing a lot of production because of the bad checkpointing and some of the other issues.
Believe it or not I think he is having problems benchmarking these proteins because of the new method that they are using to break down the folding process. There are spots in the actual folding process where the protein is just sitting there doing almost nothing. It doesn't take long to calculate nothing. Then there are times when the calculations are more intense which means that the wu's will take considerably longer even if the end frame is the same size. Kind of the same way your PC renders images. It takes little effort to produce a solid blue picture that will last for five minutes, but it takes longer for it to render a complicated moving image that will run for five minutes.