Formula Boinc Sprints 2018

[Markfw]

So what does this application do ? as in accomplish? any medical research ?

 

[[H]Coleslaw]

Astrophysics, so no medical.

 

[StefanR5R]

While camb_legacy tasks are still being generated, I am not sure if or when they will be validated.
https://www.cosmologyathome.org/forum_thread.php?id=7546
https://www.cosmologyathome.org/server_status.php (shows "script_validator (camb) Not Running")

 

[StefanR5R]

Three links to discussion threads of the current sprint:

C@H forum, "Formula BOINC sprint at Cosmology@Home, July 2018" —
OP lets the C@H admin know about the sprint. Doesn't probably matter though because (a) while Marius is friendly and responsive when he is around, he is not around very often..., and (b) the site doesn't really budge under the sprint load yet.

FB forum, "Formula Boinc Killed Cosmology@Home Today" —
You can tell from the title that the thread was opened by bcavnaugh. :-) The premise of the OP is wrong. Tasks are available all the time. It's just that people should not expect work generators to operate at faster-than-light speeds.

FB forum, "New sprint starting today 19/07" —
OP points out that the camb_legacy validator is not running, which is indeed a nasty pitfall. Asks whether the sprint project can be changed. IMO, it wouldn't be proper to switch the project after announcement.

 

[Ken g6]

Well, the race has started, and it looks like the server's really getting hammered now. I've gotten two whole tasks in the last 20 minutes.

 

[StefanR5R]

Two of my hosts still have bits in their workqueue. The others are receiving new work; not as fast yet as they can complete it, but almost as fast, and it seems to be getting better is not getting better yet.

Another eternal problem of mine is my inadequate upstream bandwidth. A single-socket Xeon completed 1000 camb_boinc2docker tasks in about 15 hours, but will apparently take the whole night to upload them. And because the client doesn't request new tasks when there are too many pending uploads, I needed to start a secondary client instance on it.

Edit,
when all of my hosts are uploading at the same time (they are now), I need almost half as much time to upload one camb_boinc2docker task as I needed to compute it.

 

[zzuupp]

Well, let's see if baby's first bunkering on both boxes works.

2 valids and 40 pending

 

[StefanR5R]

[My hosts] are receiving new work; not as fast yet as they can complete it, but almost as fast, and it seems to be getting better is not getting better yet.

7.5 hours later: Still not better.

The usual period in which everybody tries to re-fill caches should be over now. I suspect the project will not generate enough work to saturate the contributor hosts for the whole duration of the sprint.

 

[StefanR5R]

Project scientist/ admin Marius posted a brief response.

• The camb_legacy work generator was only on by accident, and was switched off again now.
  
• The camb_boinc2docker work generator is doing its thing as fast as it can; no change to be expected until the sprint is over.

 

[biodoc]

My P2 is out of work. The caches on my other 2 machines are steadily decreasing.

I had 2 tasks on my P2 that were partially complete but had odd network errors so overnight boinc was not requesting new tasks.

7/20/2018 6:10:46 AM | Cosmology@Home | Not requesting tasks: don't need (CPU: ; NVIDIA GPU: )

I aborted those 2 tasks and now boinc is requesting new tasks but I'm not getting any.

 

[StefanR5R]

Additional post from Marius:

I'm very happy to have C@H participate in the Formula BOINC sprint, however, I do wish the organizers would have contacted me earlier, then we could have built a much bigger work pool in preparation (by the time I saw this post yesterday it was too late). We could have also communicated better to the sprinters the ongoing issues with the camb_legacy app, and pointed everyone to be running only camb_boinc2docker.

To clarify / confirm what has been said, camb_boinc2docker validation is running completely smoothly. Work generation is also running at normal pace, but demand is very high so it is likely slow for clients to grab workunits. camb_legacy validation on the other hand is currently broken and has been for several weeks. For those participating in the sprint, its unlikely I can fix it in time for the Sunday deadline, unfortunately.

 

[Ken g6]

Something very strange is going on. I'm getting project backoffs with "can't resolve hostname". Earlier I got notices that the project had a new URL: "${URL_BASE}". I'm out of work on one machine, and I can't return work otherwise.

Edit: Oh, and I can't attach to the project either now.

 

[StefanR5R]

The first 18 hours:

Something very strange is going on. I'm getting project backoffs with "can't resolve hostname". Earlier I got notices that the project had a new URL: "${URL_BASE}".

The occurrence of the ${URL_BASE} glitch coincides exactly with the moment from which on Gridcoin's credits/hour inclined.

 

[crashtech]

Frickin' Gridcoin...

 

[StefanR5R]

@crashtech, I suspect what we see there is Marius trying to accommodate all those who asked for camb_legacy validation in the C@H forum, and so he seems to let four instances of his segfaulting validator wade through the backlog of old camb_legacy results.

When we remained in front of Gridcoin for several hours last night and this morning, and even kept increasing our lead over Gridcoin, I concluded from Gridcoin's prior C@H output and their C@H FB marathon standing, that we were in the lead due to the combination of low camb_boinc2docker output from the work generator, camb_legacy accidentally being generated, camb_legacy not being validated, and Gridcoin members not being aware of what goes on at a given project on a daily basis.

Meaning, our lead could go away any time if major adjustments were made at the server.

So, now that Gridcoin is out of the picture, let's see if we can remain in front of SUSA. :-)

 

[Orange Kid]

Day one of Cosmology@home
We find our selves in second due mostly I think to a flip of a switch. The turning on of the camb_legacy validator that is. Gridcoin and SUSA both have a base of consistent users here that probably have a bunch of old work just now getting validated. I don't know if we will hold off SUSA with the lack of work. I'd say if you have bunkers, turn them loose in hopes of being validated before the end.
Keep hitting that update button and hang on ya never know what might happen/


1 Gridcoin ...........................….25 ......…..4,588,681
2 TeAm AnandTech .........…...18 ............1,111,601
3 SETI.USA .....................…….15 .........….976,538
4 Czech National Team ...........12 ......…….193,181
5 Planet 3DNow! ............……...10 .........….174,570
6 Rechenkraft.net ..................….8 ..............137,734
7 BOINC@AUSTRALIA .............6 .........…..114,427
8 L'Alliance Francophone ......….4 .........…….85,118
9 UK BOINC Team ...............…..2 .........…….81,119
10 USA ..............................……..1 .........…….68,730

 

[Orange Kid]

Some stats from FreeDC for Cosmology.
These aren't getting updated much though.

xii5ku 380,089
biodoc 178,091
emoga 85,895
crashtech 75,972
Ken_g6 61,961
Orange Kid 13,112
Bremen 6,353
10esseeTony 3,472
zzuupp 1,752
Howdy2u2 162
Perry Maskell 944

 

[StefanR5R]

The second 18 hours.

When SETI.USA almost lost against us at the last sprint in NFS, they must have realized that they better not wait until Saturday to join in on the sprint.

Until a day ago I was worried about myself being unprepared for a Cosmology@Home sprint, as I did not have VirtualBox installed on my dual socket hosts. I corrected this only on Thursday evening. But it turns out that I wouldn't have benefited from getting more tasks on Thursday: My output is neither constrained by CPU power, nor by task availability. It is solely capped by my upload bandwidth. Earlier today I even did something which I normally never do during an FB sprint: I switched on another project in order to get better processor utilization. (It's a project with low network traffic, and I set it to 0 % resource share.)

So, while I read complaints about task availability in several forums, and am too seeing how my hosts struggle to receive tasks, this is not what holds back my own output.

 

[Orange Kid]

Cosmology@home day two:
Well it seems the expected has happened and we find ourselves in third. In the medals on the podium. That's a good thing not a bad thing. (sorry too much Donald lately) UK has dropped out of the points and the Russia team has moved up into the points. France and Australia are the close race to watch. If only the tasks were a bit more free flowing.

1 Gridcoin ........................…......25 ............14,508,778
2 SETI.USA .....................…......18 ..............2,416,994
3 TeAm AnandTech ..................15 .........…..1,955,586
4 Czech National Team .........…12 .................682,224
5 Planet 3DNow! ..................…..10 .........…….410,307
6 Rechenkraft.net .........…............8 ............….290,605
7 BOINC@AUSTRALIA .........…..6 .........…….264,160
8 L'Alliance Francophone ............4 ..................257,660
9 USA ...............................…........2 ............…..202,110
10 Russia Team .....................…..1 .........….....182,763

xii5ku 944,523
biodoc 241,138
crashtech 146,143
emoga 138,159
Ken_g6 130,229
Orange Kid 32,777
Bremen 13,404
Howdy2u2 11,348
10esseeTony 6,150
zzuupp 5,184
Perry Maskell 2,345
GLeeM 1,554

 

[TennesseeTony]

My apologies, for my low output. I was running the non-VM tasks, then learned of the validator issue, so switched to VM tasks......and promptly ran out on all machines, so, back to power off mode they went.

 

[zzuupp]

My apologies, for my low output. I was running the non-VM tasks, then learned of the validator issue, so switched to VM tasks......and promptly ran out on all machines, so, back to power off mode they went.

The legacy tasks that I attempted to bunker, are validating. However, they do so slowly. I've got 24 valid and 55 pending. I've had a few of the vbox/blocker tasks. They make the primegrid proth searches seem like a marathon.

 

[StefanR5R]

We entered the top ten in the C@H marathon.

The third 18 hours:

 

[StefanR5R]

Seems like people went into end spurt mode, or more hosts were switched from camb_legacy to camb_boinc2docker, because I too no longer receive camb_boinc2docker work as fast as I can upload results.

 

[StefanR5R]

We are now 7th in the C@H marathon, up from 14th IIRC.

The last 18 hours of the sprint:

 

[Orange Kid]

Cosmology@home final
An interesting finish to say the least.
Looks like we moved up with a little help from our friends. Ya, we got second with a little help from our friends.
We also move from 14th to 7th in the marathon and 6 points there.
A big thanks to all for the great finish and congrats!!!
Another sprint already this week.

1 Gridcoin ...........................….25 24,290,848
2 TeAm AnandTech .........…...18 3,703,984
3 SETI.USA ...............…......….15 3,394,565
4 Czech National Team ...........12 1,396,043
5 Rechenkraft.net .....................10 674,902
6 Planet 3DNow! ..................…..8 589,277
7 L'Alliance Francophone ......….6 455,979
8 BOINC@AUSTRALIA ......…...4 417,675
9 UK BOINC Team ....................2 397,362
10 USA ..............................….....1 338,428

xii5ku 1,629,464
biodoc 311,562
crashtech 262,480
phoenicis 237,133
Ken_g6 231,144
emoga 195,905
Orange Kid 43,248
Bremen 23,845
10esseeTony 17,395
Howdy2u2 14,936
zzuupp 12,604
Perry Maskell 7,224
GLeeM 1,554
MarcelliusIkaP 70

...and the F1 results
July 22: German Grand Prix
1 Lewis Hamilton
2 Valtteri Bottas
3 Kimi Räikkönen