Posts by tralala

1) Message boards : Rosetta@home Science : DISCUSSION of Rosetta@home Journal (4) (Message 66195)
Posted 18 May 2010 by tralala
Post:
Great to see the almost 34% increase in computing power for the project! I hope many of the new users will stick around even after CASP9 is over, as Rosetta@home is truly a great project with huge potential to help humanity.

It's probably just a fluke.

Yes, the mention of recent news on boincstats front page, the inclusion of rosetta in Seti.DE's pentathlon and the boost to recruitment has probably made no difference at all... (sigh)


Maybe it has, I'd be glad. I switched from 50/50 to 95/5 Rosetta/Non-Rosetta-projects. I'm just sceptical how sustainable the increase will be.

edit: I guess once the problems at Seti are over the TFLOPs will drop to "normal" levels.
2) Message boards : Rosetta@home Science : DISCUSSION of Rosetta@home Journal (4) (Message 66176)
Posted 17 May 2010 by tralala
Post:
Great to see the almost 34% increase in computing power for the project! I hope many of the new users will stick around even after CASP9 is over, as Rosetta@home is truly a great project with huge potential to help humanity.


It's probably just a fluke.
3) Message boards : Number crunching : Not getting any new tasks (Message 62344)
Posted 21 Jul 2009 by tralala
Post:
Time to let the backup project run. They'll fix it eventually. ;-)
4) Message boards : Number crunching : SERVER PROBLEMS. (Message 62328)
Posted 21 Jul 2009 by tralala
Post:
I have run out of work and I expect quite a few as well. Look at the TeraFLOPS estimate on the startpage and you see there is something wrong.
5) Message boards : Number crunching : Not getting any new tasks (Message 62327)
Posted 21 Jul 2009 by tralala
Post:
I have the same problem:

21.07.2009 00:35:30 rosetta@home Sending scheduler request: To fetch work.
21.07.2009 00:35:30 rosetta@home Requesting new tasks
21.07.2009 00:35:35 rosetta@home Scheduler request completed: got 0 new tasks

My version is 6.6.15.

Look at the current Teraflopcount ans you see there is something wrong.

6) Message boards : Number crunching : Memory timings.... (Message 61198)
Posted 15 May 2009 by tralala
Post:
Does RAM timings affect R@H much?

I lowered mine to 4 clocks (was on 5, which is a bit high)



No not much. Look, it was discussed in this thread, where you participated and now you open another one for a questions which has been answered there?

7) Message boards : Number crunching : Validator hanging again (Message 61172)
Posted 13 May 2009 by tralala
Post:
Well the Teraflop count on the homeopage is proof as well as my result history:

http://boinc.bakerlab.org/rosetta/results.php?userid=73828
8) Message boards : Number crunching : Problems with Minirosetta Version 1.67 (Message 61171)
Posted 13 May 2009 by tralala
Post:
Obviously a widespread problem with iRp40_S workunits:

http://boinc.bakerlab.org/rosetta/workunit.php?wuid=228722758
9) Message boards : Number crunching : 100 Teraflops (Message 61023)
Posted 5 May 2009 by tralala
Post:
IMHO the most meaningful chart about project growth rate is the following one:

http://boincstats.com/charts/chart_uk_rosetta_project_active_hosts.gif

It's a pity that it is only available for 60 days. Anyone knows a chart of active hosts for more than 60 days?
10) Message boards : Number crunching : Curious about pending tasks with a quorum of 1 (Message 60897)
Posted 29 Apr 2009 by tralala
Post:
Awarding credit is not the task of the assimilators but the validators. According to the server status page all validators are operative, however none of my results have been validated since the outage too. Maybe the validators are catching up (edit: most probably the raid and/or the polyserve file system must be rebuild) but maybe there is some "damage" left from the outage. I hope someone of the team will look into it.
11) Message boards : Number crunching : Report long-running models here (Message 60850)
Posted 27 Apr 2009 by tralala
Post:
http://boinc.bakerlab.org/rosetta/result.php?resultid=246798595
12) Message boards : Number crunching : Wild credit fluctuations (Message 60476)
Posted 3 Apr 2009 by tralala
Post:
To my knowledge, these fluctuations are caused by the credit system as discussed here.


Yeah I know that the credit system causes fluctuations and I observe them for two years now. Fluctuations are inherent in the credit granting system. However I'm not talking about the fact that there are fluctuations, but about the magnitude of them. Right now I have the feeling the fluctuations are greater than usual. I recall that in the past the credit system had a hiccup which caused unreasonable credit granting and it was fixed after it was posted. However it seems not a widespread anomality so I'm probably just a bit unlucky right now.

Another example:
http://boinc.bakerlab.org/rosetta/result.php?resultid=240534098

edit: By the way, that was me explaining the new credit system. ;-)
13) Message boards : Number crunching : Report long-running models here (Message 60433)
Posted 1 Apr 2009 by tralala
Post:
http://boinc.bakerlab.org/rosetta/result.php?resultid=239774765

19 credits? Just because this unit was more difficult to generate a decoy? Pretty rough for 12 hours work.


This is one WU for this thread
14) Message boards : Number crunching : Wild credit fluctuations (Message 60431)
Posted 1 Apr 2009 by tralala
Post:
more suspicious WUs:

http://boinc.bakerlab.org/rosetta/result.php?resultid=239888329
http://boinc.bakerlab.org/rosetta/result.php?resultid=239972090

Those are way out of line, I suggest a staff member to check the credit granting process.
15) Message boards : Number crunching : 100 Teraflops (Message 60391)
Posted 30 Mar 2009 by tralala
Post:
Well Einstein is down and Seti has not enough work to send out, so naturally there are a lot idling hosts out there. Let's hope some will stay.
16) Message boards : Number crunching : Immer weniger Credits für Rosetta-WUs ?!! (Message 60390)
Posted 30 Mar 2009 by tralala
Post:
Man bekommt hier unterdurchschnittlich Credits. Das ist aber schon eine Weile so und wird sich wohl kaum ändern, weil das Creditvergabesystem so momentan eingespielt ist.

Bei mir gibt's gerade wilde Fluktuationen in der Creditvergabe, was ich bislang zumindest bei längeren Laufzeiten und bei mir nicht beobachtet habe. Etwas stabilere Ergebnisse solltest Du aber haben, wenn Du mal Deine Laufzeit auf 4h oder 8h erhöhst, was bringt denn eine Stunde Laufzeit, wenn Du sowieso die ganze Zeit crunchst?

17) Message boards : Number crunching : Wild credit fluctuations (Message 60389)
Posted 30 Mar 2009 by tralala
Post:
For some days I observe wild fluctuations in granted credits. See here:

http://boinc.bakerlab.org/rosetta/result.php?resultid=239357351
http://boinc.bakerlab.org/rosetta/result.php?resultid=239014865
http://boinc.bakerlab.org/rosetta/result.php?resultid=239016447
http://boinc.bakerlab.org/rosetta/result.php?resultid=239181815
http://boinc.bakerlab.org/rosetta/result.php?resultid=239266781

Such fluctuations are not normal on this host. Some results however show normal variations, see my results history:

http://boinc.bakerlab.org/rosetta/results.php?userid=73828

Could be coincidence but perhaps a staff member wants to check whether the credit granting is working properly.
18) Message boards : Number crunching : So the "new" database server.... (Message 59926)
Posted 2 Mar 2009 by tralala
Post:
Is it this one?

*Database server

* dual dual-core 1.8GHz Opterons
* 16 GB RAM
* 2.6.11-1.1369_FC4 SMP kernel (64 bit)
* 212 GB RAID10 (14 X 36 GB 15K SCSI)
* dual GB NIC

Or is this the old one?


afaik this is the old one. I asked the same question in the other new-database-server-thread.
19) Message boards : Number crunching : "upgrading the database server"? (Message 59925)
Posted 2 Mar 2009 by tralala
Post:
What are the specs of the new database server?
20) Message boards : Rosetta@home Science : new supercomputer (Message 55265)
Posted 24 Aug 2008 by tralala
Post:
Actually, we are in the planning stages of moving our public automatic Rosetta server, Robetta, to BlueWaters. Our present "Robetta system" allows scientists anywhere to access Rosetta's methods but is brutally over-subscribed. A large number of scientists waiting for the service to return results have given up on us. This has distressed us but we just don't have the resources to fill the need. We have been collaborating w/ NCSA to use the cycles in some of their installed clusters to try and knock down the queue length but it has only been partially successful.

So, we're going to go big-time on this and see if we can't make BlueWaters/Robetta a nearly instantaneous Rosetta service for researchers around the globe. There's LOTS and LOTS of scaling issues to be addressed but we're looking at them and applying for $$$ to hire some folks to help w/ all of this.

Anyway, I thought it was interesting enough to post a bit about BlueWaters.... we'll keep you abreast of our progress along these lines....


I think with some improvements on low-latency implementation one could use BOINC for this very well. Turnaround times of a few hours are easily achievable, expecially if one can choose the most relieable hosts in a big pool.
I guess there are no funds for such a low-cost-approach and asking for some millions for a BlueWater-Thingie is much cooler. ;-)


Next 20



©2020 University of Washington
https://www.bakerlab.org