Newbie Q&A, if you\'re new, have a view!

Message boards : Number crunching : Newbie Q&A, if you\'re new, have a view!

To post messages, you must log in.

Previous · 1 . . . 3 · 4 · 5 · 6 · 7 · 8 · Next

AuthorMessage
Mod.Sense
Volunteer moderator
Project administrator

Send message
Joined: 22 Aug 06
Posts: 3542
Credit: 0
RAC: 0
Message 67065 - Posted: 2 Aug 2010, 16:46:24 UTC

Not sure what is happening there lnthomp. It appears you\'ve already aborted the task. That was what I was going to suggest.
Rosetta Moderator: Mod.Sense
ID: 67065 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Foxtrot

Send message
Joined: 26 Sep 09
Posts: 1
Credit: 201,816
RAC: 0
Message 67958 - Posted: 4 Oct 2010, 21:50:20 UTC

Xcuse me, but I noticed sometimes that a task which is scheduled to last for - let\'s say - 4 hours has been running for something as 50 hours (I leave one of my computers always on and maybe I don\'t look at it for some days especially when I\'m away) and is still at 40% or so. I usually stop that ones. But wasting much time. I partecipate also to other projects (Seti e WCG) and it never happens with their task. Can somebody help me? Foxtrot
ID: 67958 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Alice&Dave

Send message
Joined: 9 Oct 10
Posts: 1
Credit: 67,723
RAC: 0
Message 68499 - Posted: 9 Nov 2010, 19:56:39 UTC - in response to Message 17972.  

I wanted to create a place for all of the new participants to drop their questions and get help in getting started and with running Rosetta. The hope is that we can build a good supplement to the FAQs.

If you are a newbie, welcome to Rosetta!! Post your questions in here. If you are not a newbie, post a question and the answer here that you know the newbies will have. And try to provide links where they can get more details on the issue.

Q: Progress Percent not advancing?
A: Rosetta recomputes the progress percent at the end of each model. The model number is shown in the graphic. As long as the \"steps\" are continuing to progress, it is working. Once it completes the model it is working on, it will recompute the progress. At that point the progress % will be determine by looking at the time it took to complete the first model, as compared to your WU runtime preference. If your WU runtime preference is low (<4hrs) you will frequently see the progress % jump from 1.xx to 100%, or into the 50% range. Basically, each different protein takes a different time to crunch a model. Some proteins will crunch for several hours to complete a single model. Others will crunch a model every 5 or 10 minutes. It is the nature of the science being done with Rosetta.

BOINC and Rosetta have measures in place that will abort work units that aren\'t running properly. So, in general, unless you see some specific advice to the contrary, you should NOT abort work units (WUs).

ID: 68499 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
transient
Avatar

Send message
Joined: 30 Sep 06
Posts: 376
Credit: 9,235,093
RAC: 2,527
Message 68533 - Posted: 10 Nov 2010, 5:59:42 UTC - in response to Message 67958.  

Xcuse me, but I noticed sometimes that a task which is scheduled to last for - let\'s say - 4 hours has been running for something as 50 hours (I leave one of my computers always on and maybe I don\'t look at it for some days especially when I\'m away) and is still at 40% or so. I usually stop that ones. But wasting much time. I partecipate also to other projects (Seti e WCG) and it never happens with their task. Can somebody help me? Foxtrot


A task should not run longer than the runtime you setup in your preferences + 4 hours. You could try resetting your computer. If the task continues to show no progress (percentage does not increase), abort the task.
ID: 68533 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator
Project administrator

Send message
Joined: 22 Aug 06
Posts: 3542
Credit: 0
RAC: 0
Message 68539 - Posted: 10 Nov 2010, 16:37:46 UTC - in response to Message 68533.  

Xcuse me, but I noticed sometimes that a task which is scheduled to last for - let\'s say - 4 hours has been running for something as 50 hours (I leave one of my computers always on and maybe I don\'t look at it for some days especially when I\'m away) and is still at 40% or so. I usually stop that ones. But wasting much time. I partecipate also to other projects (Seti e WCG) and it never happens with their task. Can somebody help me? Foxtrot


A task should not run longer than the runtime you setup in your preferences + 4 hours. You could try resetting your computer. If the task continues to show no progress (percentage does not increase), abort the task.


I concur with transient, with the qualification that we are talking about actual CPU time here, not the elapsed time shown in the recent editions of BOINC Manager. If the machine is busy running something with a higher priority, the BOINC tasks may be out there for hours and not get any CPU time. That means the watchdog doesn\'t have any CPU time to see what\'s going on as well. It also probably means there is another application burning CPU cycles on your machine.
Rosetta Moderator: Mod.Sense
ID: 68539 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Hisannah

Send message
Joined: 9 Nov 10
Posts: 1
Credit: 22,124
RAC: 0
Message 68654 - Posted: 17 Nov 2010, 18:06:48 UTC

What does Communication Deferred mean?
ID: 68654 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator
Project administrator

Send message
Joined: 22 Aug 06
Posts: 3542
Credit: 0
RAC: 0
Message 68656 - Posted: 17 Nov 2010, 18:41:23 UTC

What does Communication Deferred mean?


When your machine would like to reach the project, it waits until you\'re configured preference (if any) for when to use the network allows BOINC\'s use, then it tries to contact the project. If this contact fails for any reason (network problems, no work from project, project servers down) BOINC sets a delay before trying again. This avoids having it try constantly sucking up bandwidth for you and the project, trying repeatedly to hit a server which may be down. Each time contact fails, the delays get progressively longer. This helps the project servers recover, by not having everyone hit them at the same time.
Rosetta Moderator: Mod.Sense
ID: 68656 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Rick Spies
Avatar

Send message
Joined: 5 Aug 10
Posts: 5
Credit: 89,371
RAC: 0
Message 68670 - Posted: 20 Nov 2010, 1:50:24 UTC

I\'m pretty new to Rosetta (but have been on SETI@home for >10 years).

Everything has been working just fine, but it\'s been many days since I last received work. I\'m about to run out of work units, with two in process now and only one more to replace them when completed. I do want to keep my CPUs busy. The Home Page makes it look like there are plenty of jobs available.

Any ideas? Below is today\'s messages so far:

11/19/2010 6:21:40 AM rosetta@home Restarting task Ross2X3_SAVE_ALL_OUT_r435_005_22542_1381_0 using minirosetta version 217
11/19/2010 6:21:43 AM rosetta@home Restarting task thioredoxin_LESSPCSCST_BOINC_abrelax.default.v1_SAVE_ALL_OUT_22535_3544_0 using minirosetta version 217
11/19/2010 6:21:44 AM rosetta@home Sending scheduler request: To fetch work.
11/19/2010 6:21:44 AM rosetta@home Requesting new tasks for GPU
11/19/2010 6:21:46 AM rosetta@home Scheduler request completed: got 0 new tasks
11/19/2010 12:58:34 PM rosetta@home Sending scheduler request: To fetch work.
11/19/2010 12:58:34 PM rosetta@home Requesting new tasks for CPU
11/19/2010 12:58:35 PM rosetta@home Scheduler request completed: got 1 new tasks
11/19/2010 12:58:37 PM rosetta@home Started download of 1ZGG_pcs_cst_files.r2.v1.zip
11/19/2010 12:58:43 PM rosetta@home Finished download of 1ZGG_pcs_cst_files.r2.v1.zip
11/19/2010 1:42:37 PM rosetta@home Sending scheduler request: To fetch work.
11/19/2010 1:42:37 PM rosetta@home Requesting new tasks for GPU
11/19/2010 1:42:39 PM rosetta@home Scheduler request completed: got 0 new tasks


Thanks,
Rick
=================================================
There is never a wrong time to do the right thing
ID: 68670 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator
Project administrator

Send message
Joined: 22 Aug 06
Posts: 3542
Credit: 0
RAC: 0
Message 68671 - Posted: 20 Nov 2010, 2:12:00 UTC

R@h does not have GPU tasks. The only time shown in your messages that CPU work was requested, you received some. How are your network settings configured? How often is BOINC to connect and how many additional days of work are you asking it to keep on-board?
Rosetta Moderator: Mod.Sense
ID: 68671 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Rick Spies
Avatar

Send message
Joined: 5 Aug 10
Posts: 5
Credit: 89,371
RAC: 0
Message 68677 - Posted: 21 Nov 2010, 6:10:33 UTC - in response to Message 68671.  

R@h does not have GPU tasks. The only time shown in your messages that CPU work was requested, you received some. How are your network settings configured? How often is BOINC to connect and how many additional days of work are you asking it to keep on-board?


The GPU messages are something quirky about my BOINC installation. They appear even though I\'ve got \"Use GPU...\" turned-off. This has not been a problem for either SETI@h or R@h. SETI@h can use nVidia GPUs; I have an ATI anyway.

My BOINC network configuration is completely unrestricted; it can do whatever it wants, whenever it wants, etc....

I ask for 10 days of work, my norm from participating on SETI@home, which goes down often. (It\'s been down for ~3 weeks right now.)

My last two R@h work units are being crunched now. Since it takes ~30 hours to crunch one work unit, these are almost certainly the same WUs I was running yesterday. (This PC is off a night.) In 15.5 hours, I\'ll have no work.

Today\'s log:
11/20/2010 10:18:57 AM rosetta@home Sending scheduler request: To fetch work.
11/20/2010 10:18:57 AM rosetta@home Requesting new tasks for GPU
11/20/2010 10:18:58 AM rosetta@home Scheduler request completed: got 0 new tasks
11/20/2010 10:51:30 AM rosetta@home Computation for task Ross2X3_SAVE_ALL_OUT_r435_005_22542_1381_0 finished
11/20/2010 10:51:30 AM rosetta@home Starting 1ZGG_R2_LESSPCSCST2_BOINC_abrelax.default.v1_SAVE_ALL_OUT_22545_19922_0
11/20/2010 10:51:32 AM rosetta@home Starting task 1ZGG_R2_LESSPCSCST2_BOINC_abrelax.default.v1_SAVE_ALL_OUT_22545_19922_0 using minirosetta version 217
11/20/2010 10:51:33 AM rosetta@home Started upload of Ross2X3_SAVE_ALL_OUT_r435_005_22542_1381_0_0
11/20/2010 10:51:37 AM rosetta@home Finished upload of Ross2X3_SAVE_ALL_OUT_r435_005_22542_1381_0_0
11/20/2010 10:51:38 AM rosetta@home Sending scheduler request: To report completed tasks.
11/20/2010 10:51:38 AM rosetta@home Reporting 1 completed tasks, requesting new tasks for GPU
11/20/2010 10:51:39 AM rosetta@home Scheduler request completed: got 0 new tasks
11/20/2010 11:15:11 AM rosetta@home Computation for task thioredoxin_LESSPCSCST_BOINC_abrelax.default.v1_SAVE_ALL_OUT_22535_3544_0 finished
11/20/2010 11:15:13 AM rosetta@home Started upload of thioredoxin_LESSPCSCST_BOINC_abrelax.default.v1_SAVE_ALL_OUT_22535_3544_0_0
11/20/2010 11:15:17 AM rosetta@home Finished upload of thioredoxin_LESSPCSCST_BOINC_abrelax.default.v1_SAVE_ALL_OUT_22535_3544_0_0
11/20/2010 11:15:22 AM rosetta@home Sending scheduler request: To report completed tasks.
11/20/2010 11:15:22 AM rosetta@home Reporting 1 completed tasks, requesting new tasks for CPU and GPU
11/20/2010 11:15:24 AM rosetta@home Scheduler request completed: got 1 new tasks
11/20/2010 11:15:26 AM rosetta@home Started download of 1T17_pcs_cst_files.r2.pnoe.v1.zip
11/20/2010 11:15:33 AM rosetta@home Finished download of 1T17_pcs_cst_files.r2.pnoe.v1.zip
11/20/2010 11:15:33 AM rosetta@home Starting 1T17_R2_pNOE_LESSPCSCST2_BOINC_abrelax.default.v1_SAVE_ALL_OUT_22550_20557_0
11/20/2010 11:15:34 AM rosetta@home Starting task 1T17_R2_pNOE_LESSPCSCST2_BOINC_abrelax.default.v1_SAVE_ALL_OUT_22550_20557_0 using minirosetta version 217
11/20/2010 9:36:38 PM rosetta@home Sending scheduler request: To fetch work.
11/20/2010 9:36:38 PM rosetta@home Requesting new tasks for GPU
11/20/2010 9:36:39 PM rosetta@home Scheduler request completed: got 0 new tasks

=================================================
There is never a wrong time to do the right thing
ID: 68677 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator
Project administrator

Send message
Joined: 22 Aug 06
Posts: 3542
Credit: 0
RAC: 0
Message 68678 - Posted: 21 Nov 2010, 6:51:43 UTC

Well, the discussion has been carried about many times before. All I can really say is that Rosetta is giving your CPU work when your machine is asking for it. The amount of work requested, and when it is requested is entirely up to the BOINC core client running on your machine, and the code that decides how and when to fetch work, and how to handle two projects when one is down is an area that is changing frequently. They are still really trying to achieve the right balances between enforcing your resource shares, always having enough work for all of the CPUs you\'ve configured for BOINC to use, returning work on time (i.e. before deadlines), etc.

10 day buffer for a machine that has full-time access to the internet seems like an odd combination, but you certainly should expect BOINC to deliver something much closer to your preference then you seem to be getting. Keep in mind that the maximum runtime preference you can set for Rosetta is 24hrs, if your machine is estimating 30hr completion times initially, it would tend to indicate you\'ve had a few tasks recently that ran long. Also, keep in mind that you\'ve probably now accrued quite a \"debt\" of CPU time to SETI and so BOINC is really hoping SETI is going to give you some work to help deliver the resource shares you have specified.

Another factor could be that BOINC will not request work for a project that is currently running tasks at \"high priority\" to try and avoid passing the deadline. If at one time you did have a 10 day buffer, the oldest of those tasks would be close to it\'s 10 day deadline and perhaps be causing this to occur.

Another thing that could be a factor is BOINC learning how many hours per day your machine is typically running. Although it sounds like you have a fair history of being rather consistent about that. So BOINC has probably got a reasonable idea that \"10 days\" of work for the way you use your machine is... say 120 hours of work (12 hrs per day?). So if you have tasks that will actually take 15.5 hours to complete, you actually have (in my example anyway) 31 hours of work because your machine is off half the time. Adjust the specifics for your duty cycle, but you get the idea.

My advice?? Don\'t worry about it. I am confident that BOINC will ask for, (and very likely receive) more work before you completely run out. I say \"very likely\" just because no project has work available all of the time, but your odds of R@h having work are very very high. Depending on your BOINC version, I have see it actually run out of work for 5 or 10 minutes before it actually asks for more. As I say, the folks at Berkeley are working on it.
Rosetta Moderator: Mod.Sense
ID: 68678 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Rick Spies
Avatar

Send message
Joined: 5 Aug 10
Posts: 5
Credit: 89,371
RAC: 0
Message 68683 - Posted: 22 Nov 2010, 20:57:06 UTC - in response to Message 68678.  

Well, the discussion has been carried about many times before. All I can really say is that Rosetta is giving your CPU work when your machine is asking for it. The amount of work requested, and when it is requested is entirely up to the BOINC core client running on your machine, and the code that decides how and when to fetch work, and how to handle two projects when one is down is an area that is changing frequently. They are still really trying to achieve the right balances between enforcing your resource shares, always having enough work for all of the CPUs you\'ve configured for BOINC to use, returning work on time (i.e. before deadlines), etc....


Thanks for your helpful explanation. It\'s good to know R@h is always there with more work; I\'ll stop fretting.

10 days of work is my habit because the SETI@home project is very unreliable and not always available. It goes down perhaps once each month, or currently, more than one month as they replace and rebuild a crashed server. S@h also shuts-down Tue-Thu every week. I guess that\'s the best they can do on a near-zero budget. I believe the majority of their money comes from donors like me.

I started R@h to fill-in these downtimes.
=================================================
There is never a wrong time to do the right thing
ID: 68683 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile robertmiles

Send message
Joined: 16 Jun 08
Posts: 689
Credit: 9,363,867
RAC: 4,634
Message 68791 - Posted: 10 Dec 2010, 1:47:54 UTC

Is Rosetta@home using a sufficiently recent enough version of the server software to use the feature for setting the minimum time before requesting another GPU workunit different from the minimum time before requesting another CPU workunit, for sufficiently recent versions of BOINC at the user\'s end?

I\'ve read that if it is, you can reduce the requests for GPU workunits by telling the user\'s end to wait nearly a week before requesting any more GPU workunits, if you\'re not planning to have any ready by then.
ID: 68791 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Marshmellowman

Send message
Joined: 13 Dec 10
Posts: 2
Credit: 156,289
RAC: 0
Message 68798 - Posted: 13 Dec 2010, 7:15:14 UTC

Hello,

Just installed BIONIC after being away for couple years. I am new to Rosetta. After setting up Rosetta I as shocked to see it running 2 services each using almost 400 mb each. I am running it in a Win28k x64 VM with 1536mb ram on a Hyper-V server. Rosetta is using almost 800mb which is half the ram in the VM. Is this normal?

I would like to install this on over systems but I am definitely not going to if its going to use this kind of resource and not play nice with the system.

Thanks,

Todd
ID: 68798 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator
Project administrator

Send message
Joined: 22 Aug 06
Posts: 3542
Credit: 0
RAC: 0
Message 68800 - Posted: 13 Dec 2010, 15:24:05 UTC

It is normal for one task to run per virtual CPU on the machine. You can configure the percentage of the available CPUs you would prefer in the BOINC preferences. You can also configure a preference for the percentage of each CPU\'s time you would like BOINC to run.

It is not abnormal for a task to use over 300MB of memory to run, especially on a machine with the memory to do so. If the machine had less memory, Rosetta avoids sending it tasks that are known to require more memory then average. You can configure the percentage of memory you would prefer in the BOINC preferences.

You have not defined \"playing nice\". But BOINC has added a lot of controls for you to inform it as to your own unique definition and then it will run within those parameters.
Rosetta Moderator: Mod.Sense
ID: 68800 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Marshmellowman

Send message
Joined: 13 Dec 10
Posts: 2
Credit: 156,289
RAC: 0
Message 68801 - Posted: 13 Dec 2010, 20:13:01 UTC - in response to Message 68800.  

Thanks, I was a little worried at first but I understand now. When I used BIONIC with SETI years ago it didn\'t utilize a lot of ram. I was not expecting the current memory utilization. When I reduced the ram available in bionic the processing stopped. I increased it some and 1 of the 2 processes is now running, the other shows waiting for memory - it picks up once the system is idle. Guess I just expected this to run in 30 - 50 meg of ram like the old bionic.

Not playing nice to me would mean something that takes more than a small percentage of system ram(5% -15%) on a pc that was expected to perform other tasks and this was low priory function.

Please understand I am not knocking the process. I was concerned because I want to install this on various pcs and was worried it would bring some of the older systems to their knees to the point of system crashes. However I installed Rosetta on an older dell with 2ghz and 512mb of ram today and it seems to do just fine.
ID: 68801 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator
Project administrator

Send message
Joined: 22 Aug 06
Posts: 3542
Credit: 0
RAC: 0
Message 68802 - Posted: 13 Dec 2010, 20:55:30 UTC
Last modified: 13 Dec 2010, 20:56:46 UTC

Right, when you limit the memory that BOINC is allowed to use, the only throttle is has to enforce that limit is preventing a task from running. If the machine is doing other tasks as well, it might be good to use the CPU settings to limit BOINC to 1 CPU. That way another CPU is always available for other work (although BOINC tasks run at the lowest possible priority), and there would never be a second task worth of memory to page out before something else can run.

Rosetta@home uses a substantial amount of RAM. But some other BOINC projects run on less. So, perhaps your older machines will be happier running other applications that consume less memory.
Rosetta Moderator: Mod.Sense
ID: 68802 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Murasaki
Avatar

Send message
Joined: 20 Apr 06
Posts: 303
Credit: 492,473
RAC: 116
Message 68803 - Posted: 13 Dec 2010, 21:00:10 UTC - in response to Message 68801.  
Last modified: 13 Dec 2010, 21:02:25 UTC

Guess I just expected this to run in 30 - 50 meg of ram like the old bionic.


The memory requirement varies between projects. Due to the nature of the research Rosetta is involved with it is one of the more memory-hungry projects using the BOINC system. Other projects often use much less memory per task.

If you find some of your systems just can\'t cope with the memory load you may want to try some less intensive projects. I recently put together a brief list of similar medical research projects on BOINC if you do want to try other options.

Edit:
Looks like Mod.Sense is a little quicker than me at typing.
ID: 68803 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Rockhound57

Send message
Joined: 2 Mar 11
Posts: 16
Credit: 1,181,412
RAC: 0
Message 69776 - Posted: 10 Mar 2011, 1:09:33 UTC - in response to Message 68803.  
Last modified: 10 Mar 2011, 1:47:08 UTC

Hi all; I am a newbie here, and have a few questions concerning the software schedules. First of all, I am using one of our under-utilized servers in an industrial environment. It is an HP Proliant DL-580, 8 Xeon CPU machine running at 3.0ghz, and with 8GB of RAM. During buisness hours, I am running it at 70% CPU time, and evenings and weekends it is running at 99%. It is in a very cold environment, so heating issues aren\'t a problem with running it that hard, in fact, none of the CPU\'s ever get over 115 degrees F.

I have the time schedules working correctly, but I always have to change the CPU time manually. Is there a way to automate CPU time changes along with the time schedules?

Also, when the machine is running at 70%, all of the CPU\'s run at 100% for a time, then drop to 30% for a few seconds, then back up to 100%. I guess that\'s how the software averages the CPU time, but it sure looks jerky, and I\'m not sure that it is good for the CPU\'s. Is there a way to throttle the CPU\'s more evenly so that they are at a constant 70% ? TIA
ID: 69776 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
transient
Avatar

Send message
Joined: 30 Sep 06
Posts: 376
Credit: 9,235,093
RAC: 2,527
Message 69778 - Posted: 10 Mar 2011, 6:37:10 UTC - in response to Message 69776.  


Also, when the machine is running at 70%, all of the CPU\'s run at 100% for a time, then drop to 30% for a few seconds, then back up to 100%. I guess that\'s how the software averages the CPU time, but it sure looks jerky, and I\'m not sure that it is good for the CPU\'s. Is there a way to throttle the CPU\'s more evenly so that they are at a constant 70% ? TIA


Instead of setting a percentage for usage, you could limit the number of cores the computer runs on. If we\'re talking a bout a windows machine here, you could also use a program called TThrottle. It is a bit smoother in managing a usage percentage. You can\'t set a percentage there but you can set a low enough temperature that it averages out to 70%.

This doesn\'t solve your scheduling issues. The only way I can think of managing that, is through a batch file. You can send commands to the manager through the command-line with boinccmd.exe. (on Windows)

ID: 69778 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Previous · 1 . . . 3 · 4 · 5 · 6 · 7 · 8 · Next

Message boards : Number crunching : Newbie Q&A, if you\'re new, have a view!



©2019 University of Washington
http://www.bakerlab.org