Putting up some new adaptive runs today

Thanks. Just for my understanding, where does this recurrent lack of tasks come from? Not enough molecules to analyze .. or not enough disk space on the server?

---

I would love to see HCF1 protein folding and interaction simulations to help my little boy... someday.

Most likely the scientists have to design the simulation which takes time and once completed they need to analyze the results and create and publish papers which summarize what they've learned. This is not a fast process so I would be more lenient towards expecting work at all times.

Most likely the scientists have to design the simulation which takes time and once completed they need to analyze the results and create and publish papers which summarize what they've learned. This is not a fast process so I would be more lenient towards expecting work at all times.

So your guess is lack of human resources, if I get you right..

---

I would love to see HCF1 protein folding and interaction simulations to help my little boy... someday.

Or ... leave GPUGrid at 100, and set backup projects to the "backup resource share" value, of 0, which literally means "Only use this project when work cannot be retrieved from other projects and I need work now ... and only get enough tasks to satisfy my idle resources."

0 Resource Share, is very useful.

I'll have to keep that trick in mind. For now I'll do the resource share 1 for the backup since I have my main projects set at 1000.

So your guess is lack of human resources, if I get you right..

It seems that is the case.

Or ... leave GPUGrid at 100, and set backup projects to the "backup resource share" value, of 0, which literally means "Only use this project when work cannot be retrieved from other projects and I need work now ... and only get enough tasks to satisfy my idle resources."

0 Resource Share, is very useful.

Really helpful -- thanks so much for the post.

Hm yes. It takes a while to come up with new meaningful projects or collaborations. We try not to send out simulations without a purpose. Also the analysis of the simulations takes much longer than the actual simulations. I can work on a dataset for a year (or multiple years in some cases) that was simulated in a month.

We sent out some adaptives (I work with ADRIA on a project) but I am discussing now with our collaborators if we should send out some more BNBS2 workunits.

Also the adaptives I send out now with Adria are some small tests which will lead to probably a pretty large project with lots of simulations. We do it step by step to find first which is the best method before we go crazy on the different proteins and thousands of simulations.

Theoretically I am sure there could be other groups/people wanting to simulate on GPUGRID instead of us having it idle, it is however quite complicated to organize collaborations, simulation priorities etc.

Thank you for your insight Stefan, it is always a pleasure for us when you keep us updated

Thank you .. that was a straight answer. I really appreciate that.

---

I would love to see HCF1 protein folding and interaction simulations to help my little boy... someday.

Theoretically I am sure there could be other groups/people wanting to simulate on GPUGRID instead of us having it idle, it is however quite complicated to organize collaborations, simulation priorities etc.

One more question, Stefan, if I may. Can your students possibly be involved in this kind of distributed computing? Maybe in the course of a laboratory tutorial, making their own simulations and evaluations on simple proteins, using spare power of GPUGRID. There are quite a few benefits from that approach,

- you could feed the (frequently empty) short run queue.
- you may save work time by outsourcing some minor problems to students
- you may benefit from creative thinking of students
- future researchers can familiarize with this kind of simulation and become group members occasionally

and eventually even distinguish between long runs, short runs and short "student" runs so that crunchers may choose between purely scientific and educational application. And in case the interface to GPUGRID is not yet open to students, that provides an opportunity for a diploma thesis.

---

I would love to see HCF1 protein folding and interaction simulations to help my little boy... someday.

Good idea!

Thank you .. that was a straight answer. I really appreciate that.

+ 1

Hm yeah the student thing could work although you would have to take into consideration all the mistakes they will make which would be wasted computational power. However we only teach one trimester per year so it would be a pretty short thing of ~3 weeks per year.

Hm yeah the student thing could work although you would have to take into consideration all the mistakes they will make which would be wasted computational power.

Right! On the other hand, how should they learn it otherwise if not by trial and error? If we generate some aha reactions in young minds, I don't consider that as wasted entirely.

However we only teach one trimester per year so it would be a pretty short thing of ~3 weeks per year.

Well, 3 weeks are 3 weeks... and maybe one or two gifted students want to learn more (and contribute something really valuable) in the course of a diploma thesis because the lab tutorial made them curious.

What you could also try is making one dedicated workplace in the lab with (limited) GPUGRID access generally available to students. There, they could upload their own simulations in their free time and feed the "educational" queue. And I am pretty sure, there will be quite a few people using it. Isn't it awesome to know that thousands of volunteers provide their computing power just to solve your riddle? How cool is that? :-))

As an incentive, you could award the most promising results once a year by a prize of symbolic nature. Just a deed, mentioning on the institute homepage and maybe a free movie ticket. Well, that was at least something in my own studenthood long time ago. ;-)

---

I would love to see HCF1 protein folding and interaction simulations to help my little boy... someday.

As an incentive, you could award the most promising results once a year by a prize of symbolic nature. Just a deed, mentioning on the institute homepage and maybe a free movie ticket.

Good natured as that sounds, I don't think competition belongs in research. This behavior leads to not posting something that went wrong so others can learn about it, and not posting a potential failure. Only the success ever gets published, and that's only half or less of the story.

I don't think competition belongs in research. This behavior leads to not posting something that went wrong so others can learn about it

Dont take a too narrow a view of it. From my own experience as a student (Techn. University Vienna) I think there can be many positive effects, students will likely team up to work on a particular Problem anyway. And they will certainly share information and have discussions about multiple approaches. I think that no one will do secret mongering because of just a symbolic prize.

And hey, competition has always been part of scientific research. There are scholarships and fellowships, grants and the race for honor and the best published work. That is the custom.

Well... that was just a minor detail of my proposal. Forget about the prize if you think it has a negative impact. Stefan, his colleagues and their professor have the best judgement anyway what could work at their department.

---

I would love to see HCF1 protein folding and interaction simulations to help my little boy... someday.

Oh we do make the students simulate :) The last years they simulated on some local machines we provided them with. Just not with enough computational power to make any real discoveries.

As bad as it sounds, from my personal experience the students of the Masters program are either too stressed out from too many classes or just not interested enough to do any real work. I could go on for hours about what percentage of exercises was just blatant plagiarism. The exceptions we tend to pick up in our lab, though. But I understand it, I also had too much work during the studies to sit down and do real research.

Teaching is indeed a really complicated matter, where you need to ispire, incentivate and motivate students and it requires lots of work, attention and preparation.

I'm actually not involved in the teaching anymore since I'll be likely moving lab soon (tm). But I'll mention the ideas to Gianni for this year's course.

..either too stressed out from too many classes or just not interested enough to do any real work. I could go on for hours about what percentage of exercises was just blatant plagiarism.

Maybe they would be more motivated if they can get access to the real GPUGRID supercomputer instead of a regular PC. Well, if I was a student again I'd be thrilled to have access to this mighty plaything ;-)

I'm actually not involved in the teaching anymore since I'll be likely moving lab soon (tm). But I'll mention the ideas to Gianni for this year's course.

Great! Whatever the outcome, thank you for your positive attitude.

---

I would love to see HCF1 protein folding and interaction simulations to help my little boy... someday.

To solve this continuous shortage I suggest a backup project within the GPUGrid project, which has the lowest possible priority. This could be a totally different research, I think something like protein folding (I'm not sure if the ACEMD is capable of this). My other thought is a very large molecule which take very long time to simulate, but the simulation is split up to vast number of pieces. Perhaps those who start this research, and those who will finally analyse the result will be different scientists, but if the science is the same it could be done.