Questions and Answers : Preferences : 14 days until deadline
Author | Message |
---|---|
StevenJ Send message Joined: 25 Aug 15 Posts: 2 Credit: 5,355,350 RAC: 0 |
It seems that Rosetta is only providing WU which are 14 days from deadline. Some other projects are posting WU which may not be due for almost a year. Is Rosetta that tight on volunteer CPU time? Or is Rosetta designed to work with such short lead times on WU? |
Sid Celery Send message Joined: 11 Feb 08 Posts: 2114 Credit: 41,100,175 RAC: 22,181 |
It seems that Rosetta is only providing WU which are 14 days from deadline. Some other projects are posting WU which may not be due for almost a year. Is Rosetta that tight on volunteer CPU time? Or is Rosetta designed to work with such short lead times on WU? I think it used to be 10 days so the current 14 days is more than it's historically been. WCG uses 7 or 10 days. Malaria uses just 3.5 days. I don't have any experience with other projects, but I'm surprised any use a whole year. I guess it's just what you're used to seeing. |
StevenJ Send message Joined: 25 Aug 15 Posts: 2 Credit: 5,355,350 RAC: 0 |
Sid, Thank you for the reply. I am also participating in the Climate Prediction project and many of their WU that my system runs have a deadline of Aug 2016. It seemed curious that there would be such a wide variation in WU deadlines between different projects. |
Murasaki Send message Joined: 20 Apr 06 Posts: 303 Credit: 511,418 RAC: 0 |
Climate prediction is focused on generating and analysing models of the world's climate with some models calculating up to a few centuries in the future. That type of research lends itself well to giving a big chunk of data to a particular computer and letting it run for several months. At key stages of the process the system can decide if the model is looking reasonable or if it has gone off the rails (like predicting the Earth will become hotter than the Sun). Good models advance the project significantly. Poor models are abandoned with the several months of processing providing little scientific benefit. Rosetta takes a different approach. There are a few different types of work done here, but at a basic level the aim is to find a protein structure with the lowest amount of energy needed to maintain its shape (the less energy needed, the more likely it is that the structure will be the stable shape found in nature). There are billions of potential protein structures for each chemical chain, which would take a single computer centuries or millenia to find the correct answer. Instead Rosetta asks many computers to do a set of quick and rough models of the protein and return the results in a short time frame. The scientists study the results and find a few rough predictions with low energy levels. A new set of tasks are issued focusing on these key areas and the many computers do a more detailed set of models and return the results. This cycle repeats until the scientists feel that they have a good prediction. |
Mod.Sense Volunteer moderator Send message Joined: 22 Aug 06 Posts: 4018 Credit: 0 RAC: 0 |
The default runtime of a R@h task is 6 hours. So, the project reasonably expects that if your machine will ever complete such a task, that it can do so in less than 2 weeks. And yes, if a year's worth of tasks were all in existence at the same time, that would probably max out the project databases and servers. Rosetta Moderator: Mod.Sense |
City & Country School Send message Joined: 11 Jan 06 Posts: 3 Credit: 1,499 RAC: 0 |
The default runtime of a R@h task is 6 hours. So, the project reasonably expects that if your machine will ever complete such a task, that it can do so in less than 2 weeks. And yes, if a year's worth of tasks were all in existence at the same time, that would probably max out the project databases and servers. |
City & Country School Send message Joined: 11 Jan 06 Posts: 3 Credit: 1,499 RAC: 0 |
I just got a download of a 6 hour task from R@h that had a due date of 10/10 which is two days not 14. I also just got a download of a task that is due on 10/12 or 4 days. Since I do not run my comp 24/7, if I go away for the weekend, the tasks can't get done. I have aborted the 10/10 one and will continue to abort such short times. |
City & Country School Send message Joined: 11 Jan 06 Posts: 3 Credit: 1,499 RAC: 0 |
And I got three more work units today that are due in two days. Please stop using such short time frames. |
Sid Celery Send message Joined: 11 Feb 08 Posts: 2114 Credit: 41,100,175 RAC: 22,181 |
I just got a download of a 6 hour task from R@h that had a due date of 10/10 which is two days not 14. I also just got a download of a task that is due on 10/12 or 4 days. I think the default runtime is now 8hrs, which doesn't make it any better, but during (and following up) CASP short lead times are needed by the project. Abort them if they're not suitable for your circumstances. Having said that, what other tasks or projects do you run that will prevent them running straight off? I can't see any at all. You also have the option of changing the runtime to something more suitable - runtimes aren't fixed. Change "Target CPU Runtime" on the website by selecting "Your Preferences" from the sidebar in Boinc and choosing Rosetta@home preferences |
Walter2209 Send message Joined: 21 Apr 17 Posts: 1 Credit: 45,837 RAC: 0 |
Hi, currently I figure out some issue with the lead time. Based on - the short lead time max 1- 2 weeks (sometimes 3 days) and - the long project run time (up to 10 hours) >> more than the half of the projects are not finished till the deadline ends. In other projects I have deadlines for more than 4 weeks and program run time of about 4-6 hours. My questions: Does it make sense to continue? Thanks for comments. |
Sid Celery Send message Joined: 11 Feb 08 Posts: 2114 Credit: 41,100,175 RAC: 22,181 |
Hi, currently I figure out some issue with the lead time. Having a quick look at your results, do you not have your computer on very much? You've had very few tasks and failed to complete (usually) 8 hour tasks in 8 days |
Questions and Answers :
Preferences :
14 days until deadline
©2024 University of Washington
https://www.bakerlab.org