Small update on Rosetta

[Bok]

Posted by David Baker




Believe me--we will! I'd like to welcome all former FAD participants to our project. We've been able to accomplish far more in the past two months with BOINC than would ever have been possible with our internal computing resources, but it is clear that we are still limited by total computing power. The landscapes on which we are searching for the global minimum are very large--we are trying to find a small needle in a very large haystack. My guess is that if we could increase our computational throughput by a factor of ten, many of the search problems would be solved.

Hang in there everybody--we will have the server problems ironed out very soon. We are still pretty new at this. Things should be very stable quite soon--we can get all the hardware necessary to handle increased data flow, and David K. is a real pro.



So hang in there!

Bok

[PY 222]

I sure hope they get everything ready in December as alot of people will have alot of resources to throw at this project.

[n7vxj]

Well, my boxes were completely outta work so I put them back on RC5 till they get the server straightened out!!

[Plum Ugly]

Hi all!
PY 222,I'm not sure they understand just how much computing power they will be getting.Or how much traffic the message board will be getting.

[PY 222]

Originally posted by Plum Ugly
Hi all!
PY 222,I'm not sure they understand just how much computing power they will be getting.Or how much traffic the message board will be getting.

Ahhh... Plum Ugly. My ultimate arch rival from FaD. We meet again.

So, you going to do Rosetta as well? If you are, then they will be in for some real traffic and puter power.

:cheers:

[Bok]

Well Plum,

consider, FAD has around 18000 users, who may or may not be active.

Rosetta has 5700 in the 2 months or so since it started, I think it can handle it just fine.

Some of the other boinc projects like Einstein, and especially Climate Prediction and Seti have far more users/hosts than FAD does and they handle it ok. (well, ignore seti on the other hand!)

They've already said on the Rosetta forums that they will put in any needed hardware if the demands become much higher.

I, for one, think this outage (which is their first) is a learning experience. The interaction has been good, they have been forthcoming and I expect it will be fixed very soon now.

As for message board traffic, I presume you mean just answering ?? As the actual net traffic is pretty much neglibible. Even the Free-DC servers use only 70Mb per day or so.

Bok

[Plum Ugly]

I've been getting errors from the message board stating that there was too much network traffic an couldn't connect to the web pages.Been surfing around to find the best setup for boinc in linux.


PY 222, yes we are looking for another home so when fad stops many will move here as this looks the best place for us to go.We are looking at several projects but this I think this maybe the best setup.I was looking at Folding@home,got a couple running it now.But I'd rather do something in cancer research.I need some thing that I can put it on and forget it like I did with Fad.50 systems is too many to have to moniter and work on + a 12 hr driving job.

[Bok]

PU,

I'd love to see you guys running Rosetta. It looks like it is running again right now. It's been error free for a few months until yesterday. I think the response we have had yesterday and today on the errors has been better than the vast majority of projects out there.

a few things thing to share.. boinc takes a little bit of getting used to, it certainly has it's quirks, but once you learn them it works pretty well.

The clients you download from the websites are compiled fairly generically. Look for optimized clients, these are compiled with more agressive compiler options and yield higher benchmark results.

Look into using boincview to manage many clients. Boinc has inbuilt RPC to allow remote control and boincview does a wonderful job of managing it across a network.

Bok

[Plum Ugly]

thanks for the info,I'll be running xandros(debian based) on most of the ones here. still trying to get it to run as a startup service
David poped over to the xpc boards and it made a difference in the way peeps were thinking.That shows me that the people at rosetta are willing to talk to us.made me feel better about the program it's self.Still gun shy from the UD days.
If ya got any links to setting up boinc on nix I'd shore like them.

[PCZ]

Do i sense competition :D


Plum Ugly

Starting boinc automatically shouldn't be too difficult.
On my RH based boxes i use rc.local

On debian you have to run this command

ln -s /etc/rc.local /etc/rc2.d/S99local

to get rc.local to run at startup.

just add something similar to this to /etc/rc.local

cd /boinc
./boinc -allow_remote_gui_rpc > /dev/null &
cd /


You might need to nohup it, i don't with my servers but do need to on my discless boxes.


cd /boinc
nohup ./boinc -allow_remote_gui_rpc > /dev/null &
cd /

[Jeff]

Just wanted to say hello to Mr. PU. Nice to see you again. http://www.adventurevision.com/smilies/wavey.gif
;)

[PCZ]

"Here are the steps we are or will be taking in the near future:

1. Weekly database purge and archive to keep the wu and results tables at managable sizes.
2. Adding more memory to the database server. Memory to fill out our server will be ordered today which will at least double the current size to 8gigs (we still have to check how many slots are available but an excess will be ordered regardless).
3. We are currently looking into getting a very beafy database server.
4. The size of work units will be increased (note however that they will still depend on factors like the size of the protein and what kind of prediction is being run).

We will also get any necessary hardware to handle increased demand."



;)

[Plum Ugly]

HI Jeff!
I think they're going to have to beef everything up to hold the influx of machines.