I have a comment regarding the hardware minimums.

Okay, I have a 700 MHz iBook that has been crunching for a little while, now (I've had a few of just about every kind of WU, I'd say.). It has made all of the deadlines, with no difficulty, but it only has half the recommended RAM. The longest I've seen a WU take is a little over 5 hours. This one is a part-time cruncher.

I have also set up a PII 266 MHz system (also 256 Mb of RAM), running Fedora Core 5. It hasn't turned in a WU, yet, but it is expected to, later today. It's set for 24 hours per WU and seems to be doing well so far. This one is a dedicated cruncher.

So, I have two questions. First, where do the recommendations come from? Do they assume "only-when-idle" crunching? On the iBook, I manually turn calculation on and off, since it makes pretty good progress while simultaneously running both iTunes and iPhoto. I only turn it off when I'm doing something that requires my processor's full attention. If I checked the "only-when-idle" box, I wouldn't be able to do nearly as much crunching. Still, usage habits make a big difference.

Second, the "If you fall below this, we recommend that you not run Rosetta," seems a bit steep. Technically, neither of my computers make the minimums, but they each seem to have no trouble making deadlines. Perhaps this could be changed to something like "If you are below this, Rosetta might require an inconvenient amount of CPU time," and give a second (lower) set of "If you don't have at least this, you won't be able to complete work units in time." requirements.

I might be in the minority, with such old hardware, but my computers will take any WU the project can throw at them. The day one of them can't, I'll take that one off the project. Perhaps there are others who, upon seeing the hardware recommendations, decide to attach their old PII and PIII computers to a different project, instead. I almost did, but I decided I'd give it a try. It turns out I had nothing to worry about. This might be a way to get more people into the project.

Maybe this change could be looked into? Although my RAC might be in the single digits, every clock cycle helps. "It's not the size of the computer in the project. It's the size of the project in the computer!"

---

"There are obviously many things which we do not understand, and may never be able to."
Leela (From the Mac game "Marathon", released 1995)

Well I totally agree.
We set up a user for our company's oldies (PII, K6...), and they are all under the minimal system requirements. However they are all dedicated crunshers, with a minimal Debian linux system, and never missed a deadline...

I guess the minimal requirements should be split in "categories", like a table:

                    | Part-time crunshing | 24/7 crunshing
-----------------------------------------------------------
User interaction    |      P4/512Mb       |      ??
-----------------------------------------------------------
Headless computer   |         ??          |  ~200Mhz/128Mb
-----------------------------------------------------------

Would be pretty easy to understand even for non-techies I guess.

Comments?

- trib'

---

Well I totally agree.
We set up a user for our company's oldies (PII, K6...), and they are all under the minimal system requirements. However they are all dedicated crunshers, with a minimal Debian linux system, and never missed a deadline...

I guess the minimal requirements should be split in "categories", like a table:

                    | Part-time crunshing | 24/7 crunshing
-----------------------------------------------------------
User interaction    |      P4/512Mb       |      ??
-----------------------------------------------------------
Headless computer   |         ??          |  ~200Mhz/128Mb
-----------------------------------------------------------

Would be pretty easy to understand even for non-techies I guess.

Comments?

- trib'

I think it is a great idea to make such a table, maby it will get more people on rosetta

---

Member of the Dutch Power Cows

Trying to get the world on IPv6, do you have it? check here: IPv6.RHarmsen.nl

Glad to hear that!
Now anyone would like to have a guess at what the cells marked with "??" should be filled with?

I don't own or administrate such computers, so I have no ideas what the minimals should be :(

Any ideas?

- trib'

---

I think its a good idea, but it's going to be very difficult to quantify. Basically, if the computer is a dedicated cruncher then it's fairly straight forward for the minimum RAM requirements for the different OSs, but if the computer is in use, I'd say keep the recommended minimum the same and suggest if it becomes sluggish then they should try the 'only when not in use' setting.

Also, I don't think there's any need to specify a CPU - just the RAM requirements as I'd expect any computer with 512MB RAM to be up to the job.

---

Well the computing power is important in the sense that you will be wasting your time crunshing if you cannot report in time for the workunit deadline...

But I agree, it's really hard to quantify since users can use different amounts of system ressources for different amounts of times...
The idea was just an overall guideline, keeping in mind that for us, more is always better :)

- trib'

---

I have a comment regarding the hardware minimums.

Okay, I have a 700 MHz iBook that has been crunching for a little while, now (I've had a few of just about every kind of WU, I'd say.). It has made all of the deadlines, with no difficulty, but it only has half the recommended RAM. The longest I've seen a WU take is a little over 5 hours. This one is a part-time cruncher.

I have also set up a PII 266 MHz system (also 256 Mb of RAM), running Fedora Core 5. It hasn't turned in a WU, yet, but it is expected to, later today. It's set for 24 hours per WU and seems to be doing well so far. This one is a dedicated cruncher.

So, I have two questions. First, where do the recommendations come from? Do they assume "only-when-idle" crunching? On the iBook, I manually turn calculation on and off, since it makes pretty good progress while simultaneously running both iTunes and iPhoto. I only turn it off when I'm doing something that requires my processor's full attention. If I checked the "only-when-idle" box, I wouldn't be able to do nearly as much crunching. Still, usage habits make a big difference.

Second, the "If you fall below this, we recommend that you not run Rosetta," seems a bit steep. Technically, neither of my computers make the minimums, but they each seem to have no trouble making deadlines. Perhaps this could be changed to something like "If you are below this, Rosetta might require an inconvenient amount of CPU time," and give a second (lower) set of "If you don't have at least this, you won't be able to complete work units in time." requirements.

I might be in the minority, with such old hardware, but my computers will take any WU the project can throw at them. The day one of them can't, I'll take that one off the project. Perhaps there are others who, upon seeing the hardware recommendations, decide to attach their old PII and PIII computers to a different project, instead. I almost did, but I decided I'd give it a try. It turns out I had nothing to worry about. This might be a way to get more people into the project.

Maybe this change could be looked into? Although my RAC might be in the single digits, every clock cycle helps. "It's not the size of the computer in the project. It's the size of the project in the computer!"

You are all correct that Rosetta CAN be run on systems below the minimum standards in some cases. But this will not always be the case. The minimum system requirements are not based on how fast a system should return a Work Unit. While this is important it is not the primary concern driving the requirements.

Systems that are slower and in particular systems with less memory are more prone to errors when the larger protein structures are processed. In addition, requiring systems with faster processors usually also eliminats older versions of operating system software. This is intentional, as it will help determine what resources might be available on a client system when the software is designed.

Currently the proteins are relatively small, so these systems are able to run them, even though they take more time. But during CASP, which is starting soon, the project will be running larger structures. Systems with less memory "MAY" have significant problems running them. Later when the models get still larger, the memory requirements might actually be raised still higher.

The project welcomes people to try these machines, but the idea is to point out that you will most likely experience higher error rates (in some cases very high error rates). In addition most of the current models use about 200 MB of memory when they are running. This leaves very little memory for other tasks on a system with 256MB of memory. Most of you have indicated that the machines you are running that are below the requirements are usually run as dedicated processors, or that you turn the application on and off when you use the machine for other tasks. Under those conditions you will be able to run them with some success on normal model runs, but they will be less stable than more highly configured systems. But more to the point the average user is not a dedicated cruncher.

For this reason it is unlikely that the project will encourage people to use machines that fall below the existing requirements, and there has been discussion of raising the bar still higher.

---

Moderator9
ROSETTA@home FAQ
Moderator Contact

Systems that are slower and in particular systems with less memory are more prone to errors when the larger protein structures are processed.

Aside from the client "erroring-out", is it possible to tell whether there is an error in a result my computers are returning? I know we don't have quorums to validate against, but is there some way to make sure I'm returning valid results?

Both of my machines have been used for fairly heavy processing, in the past (including 24-hour-plus compiling runs on the PII during Linux From Scratch), and they haven't given me any trouble. Still, is there any way to tell if my results are valid?

EDIT: Nevermind, I found out how to tell. Hadn't seen it on the "About this result" page...

---

"There are obviously many things which we do not understand, and may never be able to."
Leela (From the Mac game "Marathon", released 1995)

Systems that are slower and in particular systems with less memory are more prone to errors when the larger protein structures are processed.

Aside from the client "erroring-out", is it possible to tell whether there is an error in a result my computers are returning? I know we don't have quorums to validate against, but is there some way to make sure I'm returning valid results?

Both of my machines have been used for fairly heavy processing, in the past (including 24-hour-plus compiling runs on the PII during Linux From Scratch), and they haven't given me any trouble. Still, is there any way to tell if my results are valid?

EDIT: Nevermind, I found out how to tell. Hadn't seen it on the "About this result" page...

you can look at your results in 'My Account > My Computers' and then choose a computer under 'Computer ID'. Under 'results' you can look at any of the results returned. If they're not erroring out then they're almost certainly as good as anyone elses results!

HTH
Danny

---

...Aside from the client "erroring-out", is it possible to tell whether there is an error in a result my computers are returning? I know we don't have quorums to validate against, but is there some way to make sure I'm returning valid results?
...

Actually you can tell a lot from the graphic display. If you can complete a single model, and the Work Unit reports back (normally), the protein result will be useful. Even if the Work Unit errors out, if the completed models get reported back, then the protein science is valid.

But, since this project is in large part also about developing the modeling software itself, even the errors themselves are important science. This last point is often never noticed by users who are having a lot of errors.

That is why the project has instituted a weekly award of credits for error Work Units. The errors are important to the goal of the project and even though BOINC is not set up to award credit for errors, Rosetta is.

---

Moderator9
ROSETTA@home FAQ
Moderator Contact

as for older hardware, i've got a couple of PIII machines running a veriety of projects quite happily, most of the time, all they're doing is crunching, but i do use one for simple stuff like playing music etc. and it still never misses deadlines

PIII 600MHz
512MB RAM

PIII 500MHz
384MB RAM

the 500MHz machine is the one that's used quite often, but only for simple tasks

---

Want to search the BOINC Wiki, BOINCstats, or various BOINC forums from within firefox? Try the BOINC related Firefox Search Plugins