Technical News, and SETI@home daily stats for 28.08.2007.

[Rattledagger]

Sleepless in Oakland (Aug 27 2007)

Minor issues over the weekend. One night penguin (the download server) got in a snit with the network and needed to be rebooted. No big deal there, except that traffic was vastly reduced for several hours there. Of greater concern was the swelling ready-to-assimilate queue. Normally this wouldn't be that big a deal and could wait until Monday to diagnose, but this backlog left extra workunits on disk (since they have to be assimilated before they can be deleted). Add this to our lower quorums and rising results-to-send queue, and the workunit file system almost filled up! I had to halt splitting for a while to keep this from happening. I also tried adding extra assimilator processes but this didn't help.

Jeff found the problem this morning: some new assimilator code to update the "hot pix" table in the science database was doing sequential scans for row updates. A simple "update stats" on the informix table cleaned that right up quick. The "hot pix" table will be used for the near time persistency checker (yep - we're actually working on that stuff slowly but surely). The queue, and therefore the workunit storage usage, should be draining now.

Today I've been working on getting new disk volumes on line (a continuation from my last post). Not sure why I didn't know this already, but it turns out the ext3 filesystem has an 8 Terabyte limit. So we had to adjust certain plans for volume configuration until they come out with ext4. I have no time or interest in trying any other filesystems at this point.

Last night woken up around 3:00am by a nearby 2.3 earthquake and again at 3:10am by a 2.4 at the same exact location. Actually this has been an active hot spot for the past year - right at the base of the Claremont Hotel (about a mile or two away from campus). Tonight I'll be up again around the same time to catch the full lunar eclipse, or at least I'll try to be. I'm kinda wrecked.

- Matt

#____Total Work Done____Todays WD_______AWD________overtake________Team-name
01______320.752.286______589.493______500.508______impossible______SETI.USA
02______265.973.776______329.838______304.403______impossible______SETI.Germany
03______112.715.248______235.352______173.690______impossible______L'Alliance Francophone
04______108.499.738_______76.886_______80.293______impossible______BroadbandReports.com Team Starfire
05_______96.474.400______145.783______128.295______impossible______BOINC Synergy
06_______89.859.227_______87.726_______75.333______impossible______Czech National Team
07_______79.626.192_______96.400_______80.705______impossible______SETI@Netherlands
08_______63.920.343______171.892______147.567______impossible______The Knights Who Say Ni!
09_______44.398.564______-11.580_______-3.140______14.140 days______OcUK - Overclockers UK
10_______36.285.397_______51.050_______42.217______impossible______Team Art Bell
11_______36.011.251_______18.805_______28.408______impossible______Overclockers.com
12_______32.249.801______-29.041______-16.934______1.904 days______BOINC.Italy
13_______31.949.720_______35.982_______39.487______impossible______Team 2ch
14_______26.972.642_______44.582_______43.223______impossible______The Planetary Society
15_______23.088.132_______62.760_______65.377______impossible______Team MacNN
16_______17.486.581_______-3.320__________867______impossible______Ars Technica
17________4.861.069________2.175________3.152______impossible______Universe Examiners
18_______56.680.091_______82.819_______65.839______notanoption_____TeAm AnandTech
19_________-198.904_______15.018_______21.357__________9 days______Phoenix Rising
20_________-497.124_______45.555_______40.193_________12 days______SETI@Taiwan
21_______-5.364.826_______10.819_______26.917________199 days______Team Starfire World BOINC
22_______-6.291.608_______79.171_______61.855________102 days______SETI@China
23_______-9.311.532_______-6.670_______-6.123______impossible______Hewlett-Packard
24______-10.539.934______-30.702_______-6.536______impossible______PC Perspective Killer Frogs
25______-10.597.426______-17.174_______-7.642______impossible______Amateur Radio Operators
26______-10.866.379_________-386________5.840______1.861 days______Dutch Power Cows
27______-11.039.188________9.334________6.137______1.799 days______Canada
28______-15.014.658______-55.441______-42.373______impossible______Planet 3DNow!
29______-15.578.599______-39.183______-29.269______impossible______2CPU.com
30______-15.667.682______-15.572______-13.169______impossible______Team MacAddict
31______-16.380.577______-24.471______-14.453______impossible______Team NIPPON
32______-18.845.572_______-8.843______-11.714______impossible______BOINC SETI@home RUSSIA
33______-19.068.244______-11.155______-14.571______impossible______BOINC@Denmark
34______-19.145.389________7.334_______12.520______1.529 days______UK BOINC Team
35______-19.407.862______-13.967_______-6.414______impossible______Hungary
36______-20.163.011_______40.985_______57.920________348 days______BOINC@AUSTRALIA
37______-20.569.382________7.549_______11.450______1.796 days______US NAVY
38______-21.055.643______-41.618______-31.800______impossible______Portugal@Home
39______-23.415.361________3.748_______10.161______2.304 days______U.S.Air Force
40______-24.271.936______-69.028______-52.120______impossible______Picard
41______-24.600.737______-34.745______-15.935______impossible______SETI.hr
42______-25.202.367______-64.537______-49.599______impossible______LittleWhiteDog
43______-25.609.132______-30.790______-24.273______impossible______SETI@klamm.de
44______-26.634.055______-40.253______-30.843______impossible______Team EDGE
45______-27.078.365______-35.443______-27.094______impossible______HispaSeti & BOINC
46______-27.621.900_______30.188_______21.805______1.267 days______AUSTRIA - NATIONAL - TEAM
47______-29.005.438______-40.308______-32.346______impossible______SETI Sverige [Sweden]
48______-29.880.443______-16.902_______-9.448______impossible______BOINC.SK
49______-30.621.544______-49.108______-38.557______impossible______BOINC UK
50______-31.506.835______-45.726______-29.252______impossible______PurePC.pl

Appart for Anandtech's stats, it shows how much more/less than Anandtech.
Also shows based on Average Work Done how many days for Anandtech to overtake the team, or be overtaken by a team behind

 

[petrusbroder]

Thanks for the stats, Rattledagger!

I have added one more comp to the crunching. But we will be overtaken!

BTW, welcome back! I hope you had a good vacation! :sun: and that the work awaiting you when you came back was not too bad ...

 

[Assimilator1]

Ah is that why your DPAD output has dropped somewhat?

Thx RD

I hope they don't get any bad quakes their!.........

 

[petrusbroder]

Originally posted by: Assimilator1
Ah is that why your DPAD output has dropped somewhat?

Thx RD

I hope they don't get any bad quakes their!.........

No, that is not the cause ... it is because one of my DPAD-comps is out of whack ... it is getting well soon, I hope !

 

[hymy]

So you guys do still talk about seti. Well I'm finnally back to seti, after years of destitute bumbling with a celeron 300. I just wasn't going to run anything on that. But I got a good deal at tigerdirect on a refurbished athlon x2 5200+ computer. I wasn't really thinking DC at the time, but it does churn out a lot of work units (more than 10 a day).

Hopefully it will help some, I don't really understand how credit works, so I got no idea how high it's RAC will be. It's at 95 and climbing now, might go as high as 700, But I haven't seen anybody running this processor so I'm guessing based on other slower processors.

 

[Assimilator1]

Nice rig ,& good to see you about Hymy ,I think it's been a very long time!
Did you race me in one of the old SETI classic races?

Btw re BOINC ,I still don't get the scoring system ,it seems impossible to compare systems output now except using RAC which is rather long winded & unreliable :roll:
I had meant to try to 'capture' a typical WU so we could benchmark our rigs but I don't know what a typical WU is anymore & I can't remember what RD said about how to do it now

RattleDagger?

 

[Rattledagger]

Originally posted by: Assimilator1
Nice rig ,& good to see you about Hymy ,I think it's been a very long time!
Did you race me in one of the old SETI classic races?

Btw re BOINC ,I still don't get the scoring system ,it seems impossible to compare systems output now except using RAC which is rather long winded & unreliable :roll:
I had meant to try to 'capture' a typical WU so we could benchmark our rigs but I don't know what a typical WU is anymore & I can't remember what RD said about how to do it now

RattleDagger?

The most common AR with "classic" data was around 0.426, with similary-recorded data with Multibeam this would be AR around 0.256.

But, it seems much of the "classic"-data is basically the recorder just drifting, not tracking anything in particular, while for Multibeam it's more commonly tracking something, so that is the "common" AR now is more or less anyone's guess...


As for credit, there's "good" AR and "bad" AR, in theory it should be completely linear with time, but this doesn't happen in practice. Now, hopefully "soon" most of the disrepancies will be removed, but you'll still have VHAR, there some cpu's is just so much better on the particular calculations that there'll always be large differences cross-computers.

It seems the "good" credit/hour is for 0.2258 to 0.4, while 0.4 to 0.8 is mediocre, and 0.8 to 1.127 is dreadful. VHAR is either exctremely excellent or mediocre, depending on cpu...


In any case, if you're going to compare computers, using RAC has it's disadvantages, since "luck of the draw" and so on, but also how "fast" whichether you're paired with returns wu's, influences this.

So, probably better would be to calculate credit/hour, and compare this. But, make sure to include the AR, and try to compare credit/hour to the same AR on different computers...

 

[Coquito]

Originally posted by: Rattledagger

Originally posted by: Assimilator1
Nice rig ,& good to see you about Hymy ,I think it's been a very long time!
Did you race me in one of the old SETI classic races?

Btw re BOINC ,I still don't get the scoring system ,it seems impossible to compare systems output now except using RAC which is rather long winded & unreliable :roll:
I had meant to try to 'capture' a typical WU so we could benchmark our rigs but I don't know what a typical WU is anymore & I can't remember what RD said about how to do it now

RattleDagger?

The most common AR with "classic" data was around 0.426, with similary-recorded data with Multibeam this would be AR around 0.256.

But, it seems much of the "classic"-data is basically the recorder just drifting, not tracking anything in particular, while for Multibeam it's more commonly tracking something, so that is the "common" AR now is more or less anyone's guess...


As for credit, there's "good" AR and "bad" AR, in theory it should be completely linear with time, but this doesn't happen in practice. Now, hopefully "soon" most of the disrepancies will be removed, but you'll still have VHAR, there some cpu's is just so much better on the particular calculations that there'll always be large differences cross-computers.

It seems the "good" credit/hour is for 0.2258 to 0.4, while 0.4 to 0.8 is mediocre, and 0.8 to 1.127 is dreadful. VHAR is either exctremely excellent or mediocre, depending on cpu...


In any case, if you're going to compare computers, using RAC has it's disadvantages, since "luck of the draw" and so on, but also how "fast" whichether you're paired with returns wu's, influences this.

So, probably better would be to calculate credit/hour, and compare this. But, make sure to include the AR, and try to compare credit/hour to the same AR on different computers...

Thank you for the explanation. Now to make sense of it all.

Where are my crayons?

 

[hymy]

@A1

Yeah I'm certain that we did race some, I beleive you always whipped me too. Thus it's kinda nice to be outproducing you. Sad thing is I have something like 1200 credits pending.

Seems to me it wouldn't matter if it was a typical workunit or not, just need everyone to use the same one. Might even be more accurate if it was a hard one.

 

[Assimilator1]

lol
Fair point about the WU type ,though it would be better to choose at least a common WU AR.

Oh btw ,my main rig (C2D @3.2GHz) isn't on SETI atm ,but if it were that would boost my output alot
Talking of output boosts ,I got my sons rig crunching for me again ,I'd forgotten to put in the startup shortcut when I'd last messed around with it .
It won't be stacks more output as its an XPM @2.3GHz & is not on all the time ,but it'll still help.

RD
It seems the "good" credit/hour is for 0.2258 to 0.4, while 0.4 to 0.8 is mediocre, and 0.8 to 1.127 is dreadful. VHAR is either exctremely excellent or mediocre, depending on cpu...

Err ,do mean that you get a good amount of credits/hr with 0.2258-0.4 AR WUs?

How long do you think before we can have a rough idea of which AR WUs are common?

 

[Rattledagger]

Originally posted by: Assimilator1
It seems the "good" credit/hour is for 0.2258 to 0.4, while 0.4 to 0.8 is mediocre, and 0.8 to 1.127 is dreadful. VHAR is either exctremely excellent or mediocre, depending on cpu...

Err ,do mean that you get a good amount of credits/hr with 0.2258-0.4 AR WUs?

How long do you think before we can have a rough idea of which AR WUs are common?

Yes, 0.22-0.4 should give high credit/hour. Now, I've not enough data with multi-beam, and v5.27 can also have changed things some, but would guess v5.17/v5.18 + single-beam should still give some indication of credit/hour for the different "groups"...

On multibeam, a computer, that on 0.2258-0.4 gets example 10 credit/hour, would get around 8 credit/hour on 0.4-0.8, and around 7 credit/hour on 0.8 - 1.127
On VHAR (1.127+) on the other hand the effect is unpredictable, some computers will be around 8 credit/hours, while others will be around 20 credit/hours...

Then it comes to cpu-time, the longest-running is 0.2258, and cpu-time will drop as AR increases, meaning VHAR is the decidedly fastest to crunch. This also means they've got a very short deadline.

Well, VLAR should have even longer crunch-times, but just like in "classic" VLAR's seems to be few and far between, so don't know about crunch-times and credit/hour for VLAR...

The plans is to "even-out" the credit/hour-differences, but for some reason this went "missing" in v5.27, so expect v5.28 "soon". Note, this can't fix the huge differences for VHAR.

As for how long before knows that is most common, no idea...

 

[Assimilator1]

Ok thx for the info RD :thumbsup:

I guess we'll have to wait at least a few weeks before we at least a rough idea of the common AR range.