Deprecated: Function get_magic_quotes_gpc() is deprecated in /disks/centurion/b/carolyn/b/home/boincadm/projects/beta/html/inc/util.inc on line 663
Plots

Plots

Message boards : SETI@home Enhanced : Plots
Message board moderation

To post messages, you must log in.

AuthorMessage
Josef W. Segur
Volunteer tester

Send message
Joined: 14 Oct 05
Posts: 1137
Credit: 1,848,733
RAC: 0
United States
Message 1833 - Posted: 21 Jan 2006, 20:10:34 UTC

In this post Eric said:
I haven't recalibrated the Claimed Credit values because I was surprised to find out that on average our Granted Credit is about 10% higher than the CPU time derived credit.

I'll post some plots.

Eric

I'd really like to see those plots, and think others would also find them interesting.

Meanwhile, I've done some plotting too.

I finally got around to doing some analysis of a large enough set of angle ranges to get some approximate statistics on the angle range distribution. The set is derived from 1561 WUs I've crunched since August 2000, plus 6581 another participant crunched between July 2003 and September 2005. I removed two from the set before analysis, one with AR 19.310 and the other with AR 40.320; those were probably glitches which won't repeat. Obviously most of the data is from classic WUs, that makes no difference because they're the same duration and number of data points as those for seti-boinc.

Anyhow, here's a graph with three plot lines related to the set of 8140 WUs:

For those who want to see more detail, there's a 950 x 628 pixel version.

From the green % Number line, you can see that 50% of the WUs were at angle range 0.514 or smaller. But that is not really very important for enhanced.

The Red % Time line shows that about 50% of crunch time will be spent on WUs with AR less than 0.436, and that 90% of the time will be spent on WUs with 0.39 to 1.2 ARs.

The dark blue line simply shows graphically the shape of the time estimates calculated from the splitter source code. Ingleside's tables have shown samples from the same series.

I'm not sure what Eric's "on average" implies, but any comparisons obviously need to be heavily weighted to the dominant range of work.

Note: The reason my plots use a logarithmic horizontal scale is that otherwise the data is squeezed into the leftmost 1/16. The cumulative presentation provides more useful data, even though it is not as familiar as a straightforward distribution histogram. I did produce a partial linear scale histogram which shows the small peak for very low ARs, and also shows that 1553 of the WUs (19% of the total) are within the 0.42 to 0.43 range.
                                             Joe
ID: 1833 · Report as offensive
Profile Pappa
Volunteer tester
Avatar

Send message
Joined: 13 Nov 05
Posts: 1724
Credit: 3,121,901
RAC: 0
United States
Message 1849 - Posted: 22 Jan 2006, 2:03:28 UTC
Last modified: 22 Jan 2006, 2:04:55 UTC

Joe I agree!

I have not recorded AR from every workunit... I should have... Those that I have recorded could be used as a random sample for the newer workunits...

The Nice thing is that with the increase in Users we are getting Completed Quorums... I may have to go stir up some more new Beta Testers...


Al

Thanks to Paul and Friends
Please consider a Donation to the Seti Project
ID: 1849 · Report as offensive
Josef W. Segur
Volunteer tester

Send message
Joined: 14 Oct 05
Posts: 1137
Credit: 1,848,733
RAC: 0
United States
Message 1862 - Posted: 22 Jan 2006, 5:00:58 UTC - in response to Message 1849.  
Last modified: 22 Jan 2006, 5:10:25 UTC

Joe I agree!

I have not recorded AR from every workunit... I should have... Those that I have recorded could be used as a random sample for the newer workunits...

The Nice thing is that with the increase in Users we are getting Completed Quorums... I may have to go stir up some more new Beta Testers...


Al

I use BoincLogX so the AR for all WUs is automatically saved in the sah_boinc.csv file, and is shown on the monitoring display for active units. Unfortunately the history display doesn't include the AR info, the csv file has to be opened in another program.

The number of hosts grew from 1685 on January 12 to 1774 in the latest hosts.xml file. You're doing a very good job of getting more testers, thanks! I see that BOINCstats says all 1774 are "active", but I suspect that's an artifact of the xml stats not being available long enough to tell.

[edit] If we're lucky, the additional testers will help use up the work split December 5 and force Eric to make some WUs with the new splitter code.[/edit]
                                                            Joe
ID: 1862 · Report as offensive

Message boards : SETI@home Enhanced : Plots


 
©2022 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.