2019 Formula Boinc Sprints

[lane42]

Hey, where are all my points There's not a lot, but a lot more then 480

 

[TennesseeTony]

I think they showed for the challenge, but for free-DC/other sites the update period is every 12 hours at best.

 

[Orange Kid]

….and the final sprint for the year?

Rosetta@home

 

[TennesseeTony]

yeah.........

Stefan, could you share your app_config.xml somewhere for this project, based on '4Mb L3 cache per task' usage? It is pretty easy on my 3800X CPUs, just run 50% CPU utilization, or disable HT/SMT.

 

[biodoc]

Here is a description of the Rosetta Software suite. There are a large number of projects with different proteins so memory usage will depend on how many atoms there are in the proteins being studied.

Here is the home of Rosetta Commons if you'd like to learn more about the collaborative effort to develop improvements and new applications.

 

[StefanR5R]

The Formula Boinc site needed circa 25 minutes more than 28 minutes to build the starting stats. I don't think that the R@h servers are at fault for this large lag; I guess it is the FB site which is so slow.

I wonder if FB fetched the stats of all teams circa at the same time and then took so long to take them apart, or if it fetched stats of different teams at wildly different times.

Unfortunately, I didn't capture our canonical team stats at boinc.bakerlab.org myself to find out whether it's one or the other.

 

[Orange Kid]

They are not updating.

 

[StefanR5R]

Better hold on to your bunkers still.

 

[StefanR5R]

It updated to non-zero after 22:25 UTC / before 22:30 UTC, seemingly for all teams at once.

 

[StefanR5R]

formula-boinc.org's stats updates still happen after 25 minutes/ before 30 minutes after the hour (for all teams).

 

[biodoc]

Maybe the 'start' button was pushed a little late.

 

[StefanR5R]

I took a look at Rosetta@home tasks which my 4c/8t Haswell Xeon E3 completed before the sprint, using 50 % CPUs for Rosetta@home, 50 % for desktop duties. My following two findings confirm what I remembered from earlier occasions when I ran Rosetta@home. Sample size: 30 tasks, set to 8 h target runtime.

a) 2 tasks had 7 h runtime, 3 tasks had 4...5 h runtime. Their credits/hour (runtime or CPU time) was in line with the 25 other tasks which met the target runtime of 8 h.

b) Credits/hour per task vary a lot.

minimum ............. 22 credits/h​

10th percentile .... 34 credits/h​

average ............... 52 credits/h​

90th percentile .... 67 credits/h​

maximum ............ 84 credits/h​

coefficient of variation ... 26 %​

 

[Orange Kid]

Rosetta@ home day one.
A very respectable third place for us. Hanging tough with the Big 2. Both LAF (11th) and SG (14th) are off the map in this one at the moment.
Our over all fifth is looking pretty good.



From FreeDC

No update on Rosetta at this time and tomorrow doesn't look good either.

On the FreeDC site;
Downtime Saturday Nov 30th while I get a fiber gig connection hooked up.
Will have to change all DNS once complete which will take some time to propagate.
Fingers crossed that no issues with port 80/443 inbound!!

 

[Orange Kid]

Day two Rosetta@home
Third place and looking good




From FreeDC

No update yet today, but yesterdays are there.

 

[crashtech]

That's a pretty strong 3rd against XS!

 

[Markfw]

That's a pretty strong 3rd against XS!

We actually could get 2nd, if we had a few more cores !

 

[crashtech]

Fun to think we are making them sweat a little, haha!

 

[VietOZ]

That's a pretty strong 3rd against XS!

more like we're holding on our dear pants At the current rate, we won't be able to hold 2nd.
@phoenicis would you just drop the damn bomb so I can turn off the machines for yearly cleaning?
Nice output guys! Cheers!

 

[StefanR5R]

The stats updates are getting further behind, but they are not consistent. The 05:00 update happened some at some point between 05:45 and 05:55 UTC. The 06:00 update happened earlier than 06:45 UTC.

Edit: The 08:00 update happened after 08:45/ before 08:55 UTC too. Maybe the updates will roll over into the next hour eventually.

Edit 2: The 09:00 update happened after 09:55/ before 10:05 UTC. :-|

 

[TennesseeTony]

If you haven't already, you may want to go to the webpage, sign in, and change your Rosetta run time to 1H, to get those last tasks in before the deadline. Don't forget to switch back to a longer period of time later, for slightly better ppd after the race.

 

[Markfw]

If you haven't already, you may want to go to the webpage, sign in, and change your Rosetta run time to 1H, to get those last tasks in before the deadline. Don't forget to switch back to a longer period of time later, for slightly better ppd after the race.

Where is that ?

 

[Ken g6]

Where is that ?

I mentioned this before in another place.

Adjust runtime by editing preferences at https://boinc.bakerlab.org/rosetta/prefs.php?subset=project

 

[Markfw]

I mentioned this before in another place.

How do you navigate to there, I don't see preferences anywhere in a drop down.

 

[StefanR5R]

From Ken g6's link, go to "Edit preferences". Then you get to the actual switches and menus, including the one for "Target CPU run time".

 

[Markfw]

From Ken g6's link, go to "Edit preferences". Then you get to the actual switches and menus, including the one for "Target CPU run time".

I must be blind

Exdit: I didn't realize that Kens link would work for me, but it did. I don't know how to get there from the main window, but its set to one hour now.

Thanks !