Hello,
since I registered at rosetta (may 2006) I've run every WU for 6 hours, I recently put it back to the 3 hour default.
Seeing the new graphic of the results I've asked myself if my 3 h/WU contribution is even worth? Because it'll never rach the best prediction and my result isn't used at all as far as I know.
Does it help if I switch it to 24h/WU from a scientific point of view?

thanks

---

Hello,
since I registered at rosetta (may 2006) I've run every WU for 6 hours, I recently put it back to the 3 hour default.
Seeing the new graphic of the results I've asked myself if my 3 h/WU contribution is even worth? Because it'll never rach the best prediction and my result isn't used at all as far as I know.
Does it help if I switch it to 24h/WU from a scientific point of view?

thanks

If you go and look at your results returned (hit participants in the menu at the top or click on you name) and into the computer etc..
For example this latest one on the Turion running WinXP64
https://boinc.bakerlab.org/rosetta/result.php?resultid=40924186
You will see
# DONE :: 1 starting structures built 21 (nstruct) times
# This process generated 21 decoys from 21 attempts in the information.
That means you have actually returned 21 'results' to rosetta@home in that one 'task'(aka work unit)
'nuf said as they say ;-)

---

Team mauisun.org

Are all those results used or just the lowest energy structure?
I doubt I'll ever get the lowest energy in 3 hours.

---

Are all those results used or just the lowest energy structure?
I doubt I'll ever get the lowest energy in 3 hours.

each one of those 21 decoys that your computer produced in the example FC has shown are as likely to be the low energy structure as any decoys returned, so the duration makes no difference to the quality of the decoys - the results don't get better with each decoy produced, this itteration only happens within each decoy.

HTH
Danny

---

Tom:
Since 24 hr WUs were allowed, I've been running with 24 hr time limits. I, too, with a similar RAC and Credit score, have never been the lowest RMSD, or lowest energy producer.

Switching to a 24 hr WU will give you more decoys for whatever WU you're working on. While that will give you a better chance of getting one of the best scores for that particular WU - you're just as likely to get the lowest RMSD or lowest energy award for some WU you crunched that day, no matter how you split up your time amongst WUs. The science doesn't care how you split up your time; but using longer times reduces the bandwidth used by the server and your machine. If your machine is on 8-24 hours a day, the recommendation was for 8 hours. I opted for using even less bandwidth than that.

You and I may never get the lowest RMSD or lowest energy score. We can improve our chances by improving the speed of our crunching machine.. add more crunching machines, and/or run the project longer. But when the project needs 10,000 or 100,000 or over 2 million; each of the 9,998; 99,998, or 1,999,998 non best decoys are neccessary and help the science. It'll be nice to see them come up with new approaches that will get the same accuracy with fewer decoys. One of the approaches that was described prior to CASP was going to use a run of 10,000 decoys/models to get detailed description of the landscape of the protein, and then create a new WU with that information and run 10,000 more decoys. The results of those 20,000 decoys was supposed to be as accurate as running 100,000 models/decoys from the original WU.

We may not get the Employee of the Day star, but our contributions are still valuable and neccessary.

---

Tom:
Since 24 hr WUs were allowed, I've been running with 24 hr time limits. I, too, with a similar RAC and Credit score, have never been the lowest RMSD, or lowest energy producer.

Switching to a 24 hr WU will give you more decoys for whatever WU you're working on. While that will give you a better chance of getting one of the best scores for that particular WU - you're just as likely to get the lowest RMSD or lowest energy award for some WU you crunched that day, no matter how you split up your time amongst WUs. The science doesn't care how you split up your time; but using longer times reduces the bandwidth used by the server and your machine. If your machine is on 8-24 hours a day, the recommendation was for 8 hours. I opted for using even less bandwidth than that.

You and I may never get the lowest RMSD or lowest energy score. We can improve our chances by improving the speed of our crunching machine.. add more crunching machines, and/or run the project longer. But when the project needs 10,000 or 100,000 or over 2 million; each of the 9,998; 99,998, or 1,999,998 non best decoys are neccessary and help the science. It'll be nice to see them come up with new approaches that will get the same accuracy with fewer decoys. One of the approaches that was described prior to CASP was going to use a run of 10,000 decoys/models to get detailed description of the landscape of the protein, and then create a new WU with that information and run 10,000 more decoys. The results of those 20,000 decoys was supposed to be as accurate as running 100,000 models/decoys from the original WU.

We may not get the Employee of the Day star, but our contributions are still valuable and neccessary.

Thanks for your answer, it was very enlighting!

---

I was just looking at the new results perusal service and discovered that in wu 1louA_BOINC_FILTERS_ABRELAX_SAVE_ALL_OUT_flatss0.33__1254, I had the lowest RMSD. That from just 34 predictions.

My RAC is actually a little lower then yours. Even the little guys can score sometimes!

---

Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

i never had more than 4 wu waiting...... now ive got a hundred .......whats up

---

Hello Rochester! (I'm from the OTHER Rochester, MN).

Actually, you've got about 50 WUs that haven't been reported yet. And you seem to be running them with a 3 hour time preference (that is the default in the Rosetta Preferences). That would be about 5 or 6 days of work if you are crunching fulltime. This would be a reasonable thing for BOINC to do if:

1) you are indeed crunching most all of the time, and crunching Rosetta only.
2) you have changed your General Preferences to indicate that you'd like to connect to the network about every 6 days.

Did you do that? The default there is .1 days of work (2.4hrs). So if you just changed that... then this is why you got a pile of work. If indeed you are crunching fulltime, then not to worry, the WUs have a 10 day deadline, and you should complete them all on time without any trouble.

If you do not crunch Rosetta full time, or your machine is not powered on all the time and you feel you will NOT be able to complete all of the work... go in to your Rosetta preferences and change the target runtime down to an hour (many will run longer then that, some for 4-6 hrs, but most will run in less then the current 3hr target runtime). Then update to the project. This is better then aborting WUs, because you actually do some work on each and report them back normally. Then once the pile is worked down, gradually increase the target runtime to your desired level.

I for one, like have a concise list of WUs, so I worked my way up to 24hr target runtime... but if you did that today, you'd have like 50 days of work. That's why you change gradually, so BOINC has time to adapt to the longer runtimes. This causes it to ask for fewer and fewer WUs as time goes forward.

What you saw here is "normal" if indeed you took your preference from .1 days of work on-hand ("connect to network every... days", called your cache of work) up to 5 or 6 days. And it is for this reason that I ALSO recommend to make changes to this value gradually over time... because when you jump around like that, the results can be a little unnerving.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

Hello Rochester! (I'm from the OTHER Rochester, MN).

Actually, you've got about 50 WUs that haven't been reported yet. And you seem to be running them with a 3 hour time preference (that is the default in the Rosetta Preferences). That would be about 5 or 6 days of work if you are crunching fulltime. This would be a reasonable thing for BOINC to do if:

1) you are indeed crunching most all of the time, and crunching Rosetta only.
2) you have changed your General Preferences to indicate that you'd like to connect to the network about every 6 days.

Did you do that? The default there is .1 days of work (2.4hrs). So if you just changed that... then this is why you got a pile of work. If indeed you are crunching fulltime, then not to worry, the WUs have a 10 day deadline, and you should complete them all on time without any trouble.

If you do not crunch Rosetta full time, or your machine is not powered on all the time and you feel you will NOT be able to complete all of the work... go in to your Rosetta preferences and change the target runtime down to an hour (many will run longer then that, some for 4-6 hrs, but most will run in less then the current 3hr target runtime). Then update to the project. This is better then aborting WUs, because you actually do some work on each and report them back normally. Then once the pile is worked down, gradually increase the target runtime to your desired level.

I for one, like have a concise list of WUs, so I worked my way up to 24hr target runtime... but if you did that today, you'd have like 50 days of work. That's why you change gradually, so BOINC has time to adapt to the longer runtimes. This causes it to ask for fewer and fewer WUs as time goes forward.

What you saw here is "normal" if indeed you took your preference from .1 days of work on-hand ("connect to network every... days", called your cache of work) up to 5 or 6 days. And it is for this reason that I ALSO recommend to make changes to this value gradually over time... because when you jump around like that, the results can be a little unnerving.

i set the connect from 9 days to 6 now. i only run rosetta and the computer stays on 24/7 i have it set to discontinue when i use the cpmputer and start again 10 min after the comput is idle

---

i have it set to discontinue when i use the cpmputer and start again 10 min after the comput is idle

It should be fine to run it even while the computer is in use. BOINC will only take unused processing power and release whatever other applications request. Most applications barely use any processing power, so it should get a lot even while you work - unless you do movie encoding, in which case BOINC will just idle until you're done ;)

---