The Outsider's Inside View post#013 - Up, up and Away


log in

Advanced search

Message boards : SETI@home Staff Blog : The Outsider's Inside View post#013 - Up, up and Away

Author Message
Profile KevinDouglasPhD
Project scientist
Avatar
Send message
Joined: 6 Feb 06
Posts: 107
Credit: 23,981
RAC: 0
Canada
Message 547352 - Posted: 16 Apr 2007, 21:45:28 UTC

I felt like calling this week's post Death and Taxes, but I really didn't feel like talking about either. I got my US taxes done long ago, but I haven't done my Canadian ones yet. I have until April 30 to get them done.

I've been making lots of progress on my GALFA data, though I'm driving ewen into the ground in the process. I'm trying to combine 167 days of drift scan data with 25 days of basketweave (or meridian nodding, if you like) scans, so it's taking quite a long time. Plus trying to make some preliminary datacubes, as well as all the SETHI crunching. At every step I'm gaining new insights to the data reduction process though, so it's all good. My 113th SETHI datacube finished last night, and I hope a few more will get done this week.

We had a group meeting a short while ago, though nothing groundbreaking was discussed. We talked about the quorum change and how we have just barely enough work to keep up with the new setting. There was some talk about changing from CVS to something called Subversion, though all that stuff is comp sci so I just tone it out. We will be trying to get back lapsed users with an automated email that will be best to people who haven't crunched data recently. I'm sure I'll get one of those soon enough.

The NHL playoffs have been terrific so far. Even though I picked Detroit to beat Calgary in a hockey pool, I wish the Flames were giving the Red Wings a better challenge. All the other series have been thrilling, though.

This Saturday is CalDay at UC Berkeley. Dan's giving a SETI talk on campus, but it doesn't look like anyone has the time to hang out in the SETI@home lab, so we probably won't have much of a presence that day. Too bad since we're responsible for such a large fraction of donations the university receives.

I talked to a friend at Arecibo about getting some pictures of the dish/platform during the painting project. He's looking into it...

Profile ML1
Volunteer tester
Send message
Joined: 25 Nov 01
Posts: 7939
Credit: 4,007,416
RAC: 761
United Kingdom
Message 547370 - Posted: 16 Apr 2007, 22:30:11 UTC - in response to Message 547352.
Last modified: 16 Apr 2007, 22:31:06 UTC

... There was some talk about changing from CVS to something called Subversion...

If a change is to be made there, why not jump all the way over to Git?

Regards,
Martin

____________
See new freedom: Mageia4
Linux Voice See & try out your OS Freedom!
The Future is what We make IT (GPLv3)

Profile KevinDouglasPhD
Project scientist
Avatar
Send message
Joined: 6 Feb 06
Posts: 107
Credit: 23,981
RAC: 0
Canada
Message 549105 - Posted: 19 Apr 2007, 18:10:31 UTC - in response to Message 547352.

This morning I got kicked off of using sidious, again, since my job was slowing down the boinc db server. This made me feel like taking stock of the machines available to me for my HI processing, and for no good reason I thought I'd post it here. Not a lot of quantitative detail, but I'm sure such info is obtainable elsewhere.

ewen - reliable, if a little slow. Queries to the SETHI database (stored on ewen) seem to be one reason other tasks take a while to complete on ewen. Loads of disk space, which is needed for the GALFA data I'm processing. Best thing about ewen is that large files are written to disk in seconds, rather than in minutes as is the case when using some other machines across the network. Currently running one job that makes GALFA datacubes, and one that loads GALFA spectra into a file so that gain calibrations can be made by comparing the HI line at the same position from two different days of observations. Both are running about 4 times slower than identical jobs on thumper (see below).

lando - lots of cpu, not a lot of RAM. Good for making SETHI cubes. I've got 9 such jobs going on it right now. Writes to ewen pretty quickly.

bruno - pretty fast, though low on RAM. I'm running a job that makes beamfiles for new spectra added to the SETHI database. Also a spectrum-loading job like the one described above on ewen. Both are running pretty well, though writing outputs to disk across the network seems to be the slowest part of the process. Also it's a 32 bit machine, unlike the others on this list. Still, pretty reliable - thanks donors!

thumper - a great machine, plenty of CPU and RAM. Very fast at processing data, but very slow at writing outputs to ewen. Central to boinc/sql stuff for SETI@home, so unfortunately (for me) my use of it is limited.

sidious - very powerful machine. Loads of CPU, loads of RAM, writes to ewen rather quickly. Probably the only machine we currently have with the necessary RAM to produce the GALFA datacubes we plan to release as official data products. Unfortunately (for everyone in the Galactic astronomy community) it's off limits.

Profile Dr. C.E.T.I.
Avatar
Send message
Joined: 29 Feb 00
Posts: 15988
Credit: 682,934
RAC: 114
United States
Message 549407 - Posted: 20 Apr 2007, 5:33:26 UTC - in response to Message 549105.

This morning I got kicked off of using sidious, again, since my job was slowing down the boinc db server. This made me feel like taking stock of the machines available to me for my HI processing, and for no good reason I thought I'd post it here. Not a lot of quantitative detail, but I'm sure such info is obtainable elsewhere.

ewen - reliable, if a little slow. Queries to the SETHI database (stored on ewen) seem to be one reason other tasks take a while to complete on ewen. Loads of disk space, which is needed for the GALFA data I'm processing. Best thing about ewen is that large files are written to disk in seconds, rather than in minutes as is the case when using some other machines across the network. Currently running one job that makes GALFA datacubes, and one that loads GALFA spectra into a file so that gain calibrations can be made by comparing the HI line at the same position from two different days of observations. Both are running about 4 times slower than identical jobs on thumper (see below).

lando - lots of cpu, not a lot of RAM. Good for making SETHI cubes. I've got 9 such jobs going on it right now. Writes to ewen pretty quickly.

bruno - pretty fast, though low on RAM. I'm running a job that makes beamfiles for new spectra added to the SETHI database. Also a spectrum-loading job like the one described above on ewen. Both are running pretty well, though writing outputs to disk across the network seems to be the slowest part of the process. Also it's a 32 bit machine, unlike the others on this list. Still, pretty reliable - thanks donors!

thumper - a great machine, plenty of CPU and RAM. Very fast at processing data, but very slow at writing outputs to ewen. Central to boinc/sql stuff for SETI@home, so unfortunately (for me) my use of it is limited.

sidious - very powerful machine. Loads of CPU, loads of RAM, writes to ewen rather quickly. Probably the only machine we currently have with the necessary RAM to produce the GALFA datacubes we plan to release as official data products. Unfortunately (for everyone in the Galactic astronomy community) it's off limits.


Thanks very much for that Information Kevin - good to keep the updates Sir!!!


____________
BOINC Wiki . . .

Science Status Page . . .

Profile alphax
Volunteer tester
Send message
Joined: 17 May 99
Posts: 74
Credit: 1,266,810
RAC: 0
United States
Message 550335 - Posted: 21 Apr 2007, 17:22:06 UTC - in response to Message 549105.

sidious - very powerful machine. Loads of CPU, loads of RAM, writes to ewen rather quickly. Probably the only machine we currently have with the necessary RAM to produce the GALFA datacubes we plan to release as official data products. Unfortunately (for everyone in the Galactic astronomy community) it's off limits.


Why? Is it because any jobs would "slow down the boinc db server" as you stated earlier, or is it something about the way sidious is funded or designated?

____________

Profile KevinDouglasPhD
Project scientist
Avatar
Send message
Joined: 6 Feb 06
Posts: 107
Credit: 23,981
RAC: 0
Canada
Message 551232 - Posted: 22 Apr 2007, 23:00:51 UTC - in response to Message 550335.

sidious - very powerful machine. Loads of CPU, loads of RAM, writes to ewen rather quickly. Probably the only machine we currently have with the necessary RAM to produce the GALFA datacubes we plan to release as official data products. Unfortunately (for everyone in the Galactic astronomy community) it's off limits.


Why? Is it because any jobs would "slow down the boinc db server" as you stated earlier, or is it something about the way sidious is funded or designated?


It's because my jobs would interfere with sidious's main function, to the point that I would increase the number of "slow requests" by a factor of something like 1000. I think sidious could handle it if I ran 1 or 2 SETHI datacubes, but not the ones we envision for GALFA.

I'm 5 for 6 so far for the first round of the NHL playoffs. I thought Tampa Bay would beat New Jersey, but today they came up a bit short. I still hope Vancouver beats Dallas, and even though I picked Detroit to beat Calgary I won't mind being wrong on that one. But after the way the Flames closed out game 5, I'm not sure they deserve to move on. My theory on Jamie McLennan's slash was that he was thinking he was doing it on behalf of all the goalies who've been getting run over during the playoffs. Still it doesn't excuse such a vicious play.

Message boards : SETI@home Staff Blog : The Outsider's Inside View post#013 - Up, up and Away

Copyright © 2014 University of California