Formula Boinc Sprints 2018

[StefanR5R]

As soon as you turn your look to performance per Watt, either line of desktop processors is no longer reasonable, IMO. All of the desktop processors are clocked far too high (i.e. are operated far away from the sweet spot), and have far too low processing capacity per socket (thus have relatively high platform power overhead).

--------

Asteroids@home did not issue new tasks for ~15 hours straight now. This doesn't bode well for the sprint. But it's no big surprise after having seen how this project performed at the Pentathlon.

 

[StefanR5R]

Oh wait --- the same happened in last year's FB sprint at Asteroids too.
http://asteroidsathome.net/boinc/forum_thread.php?id=611

 

[zzuupp]

Oh wait --- the same happened in last year's FB sprint at Asteroids too.
http://asteroidsathome.net/boinc/forum_thread.php?id=611

I guess I shouldn't have been surprised.

If I can get a task or two, I'll crunch 'em.

Otherwise, Good Luck TeAm!

 

[StefanR5R]

Yesterday's and today's server status is showing ~320,000 tasks in progress. This seems to be Asteroid's current capacity.

IIRC, they tweaked the server to allow ~1,200,0000 tasks in progress during Pentathlon.

I don't remember what the number of tasks in progress were during the 2017 Formula BOINC sprint Asteroids. Gridcoin's activity wasn't as high in 2017: Their entire 2017 output was 2.6 B, whereas their current 2018 output is already 3.3 B.

This means: The 2018 sprint will differ from the 2017 sprint

• by an unknown difference in project-wide limit on tasks in progress (according to the admins: imposed by the server's disk capacity),
• by Gridcoin vacuuming up about 2.5 times as many percent of the tasks-in-progress as they did last year.

@zzuupp, your getting tasks depends on (a) your requesting new tasks and (b) others and their wingmen reporting completed tasks, such that the server can recycle the disk space of these tasks.

 

[StefanR5R]

Yesterday's and today's server status is showing ~320,000 tasks in progress. This seems to be Asteroid's current capacity.

Now 290,000 and falling.

Also, next week's sprint project has been determined: It's SZTAKI Desktop Grid. Enjoy.

/s

 

[crashtech]

A few of my machines have plenty, some not enough, and a few none at all...

 

[Ken g6]

Well, I dumped some results, and it looks like someone else did too, because we've jumped ahead of SETI.USA.

 

[StefanR5R]

and it looks like someone else did too

Or all your wingmen were on our team. :-)

 

[StefanR5R]

The first 18 hours.

 

[biodoc]

That's a fine looking graph! Nice job all!

 

[TennesseeTony]

I am getting both jack and shi* on my end, regarding Assteroid tasks. I have witnessed ONE task on my machines.

Oh, and Stefan, nice sarcasm regarding next Sprint being Sztaki!

 

[StefanR5R]

From Asteroids@home's server status (daytimes are European):

• during the entire Thursday: 320,000 tasks in progress
• Friday noon: 290,000 tasks in progress
• Friday night: 240,000 tasks in progress

So on Thursday, while a whole lot of people bunkered, the number remained constant, which makes sense. On Friday, when most people went online again, the number steadily declined by a rate of 80,000 per day.

In addition, tasks where available on Thursday only before ~08:00 AM UTC, and on the entire Friday at a very low rate. IOW while tasks in progress declined, tasks ready to send did not increase but remained near zero.

Conclusions:

• If this goes on, there will not be many tasks in progress left towards the end of the sprint.
• If it is true that the server is set up to limit the rate of generated work according to disk space left, then the disk space is steadily decreasing.
  
• Yet results are being returned, tasks validated, WUs completed. Hence, the disk space is most likely being eaten up by the results.

Earlier I thought that the disk space reservations for tasks in progress were the cause for the disk space problems. But now it rather looks to me that the absence of administration, and lack of effective automated housekeeping, are Asteroids@home's problem.

They mentioned that they want to deploy new server hardware in Summer. All of the above lets me think that new hardware will at best paper over their problem in the short term, but not actually solve it. (Except if the new hardware also allows them to implement effective automated housekeeping.)

 

[Orange Kid]

Day one--Asteroids@home
With the lack of work we are amazingly in second ,but not by much.
We are again in 4th over all in the marathon.
Lets hope the lack of work is distributed evenly and we can pull out another silver.

1 Gridcoin ............…............…25 ............13,900,800
2 TeAm AnandTech ............….18 ...…......5,730,720
3 SETI.USA .............................15 ….........5,517,600
4 UK BOINC Team .........….....12 .........….1,823,040
5 Czech National Team …...….10 .........….1,760,640
6 Rechenkraft.net ..................….8 ............….996,480
7 SETI.Germany ........................6 .........…….889,920
8 Planet 3DNow! ........................4 ............….792,960
9 L'Alliance Francophone ...........2 ............….648,480
10 SETIKAH@KOREA ......….....1 .................501,120

 

[StefanR5R]

The second 18 hours.

PS,
server status: Tasks ready to send remain near 0, tasks in progress are at 200,000 and falling by 80,000/day still.

 

[GLeeM]

I thought part of FB was to encourage people to check out other projects to see how good they are. What do the Sprint choosers have against Asteroids?

 

[StefanR5R]

Several of the sprint projects in the second half of 2017 already had a capped pace of the work generator, such that the sprinters couldn't be fully supplied. This year, participation in the sprints has intensified, and at the same time the capacity in Gridcoin pools has increased. Therefore I expect bottlenecks at even more sprint projects this entire year. :-(

 

[StefanR5R]

In July 2017, we finished the Asteroids sprint with 7.9 M points (1st in league 2, 3rd overall behind GRC and CNT, 0.08 M points in front of SUSA). Links: our forum, FB's stats.

This year, we reached 7.9 M points already on Saturday, 14:00 UTC, which is after 40 of 72 hours of the sprint (or after 64 of 96 hours, including the 24 hours between announcement and start). And at that time, we were 1.0 M ahead of SUSA and 5.3 M ahead of CNT (but behind the two top teams of league 2).

Gridcoin reached their 2017 finish points already after Saturday, 10:00 UTC, i.e. after 36 of 72 hours.

With this increase of production at GRC, XS, [H], and ourselves, compared to 2017, it is no wonder that new tasks are hard to come by. Last year, @Kiska observed a rate of 6 tasks generated per second.

 

[lane42]

Just my 2 cent's,
but maybe it's time to change the way the sprints run.
Any project will be hammered with just a 24 hour notice, so do away with it.
Why not list the project's at the start of the year giving each project time to maybe lift the
cap or whatever it is for 3 or 5 day's. I know folk's don't like change, But.

 

[StefanR5R]

There really should be coordination with the project admins; at least it should be attempted. Granted, this would be work for the FB organizer and for the admins.

But open announcements of the sprint projects at the beginning of the year would cause extreme bunkering.

The SETI Wow! Event is a contest of which the project and the date is known from the beginning of the year. There was one user who bunkered for 2 months (IIRC) for this event last year. (You have to admire the work he put into this.)

Same with WCG Birthday Challenge: Bunkering begins 10 days before the race, which is the task deadline of several of WCG's subprojects.

There is a subset of DC'ers who hate Pentathlon for its 3...5 days bunkering periods. (Personally, I like bunkering. It lets me transform personal time that I invest into this hobby, as well as the basic skill that is needed for it, into extra points in contests; and I think this sort of trade is fair.)

 

[lane42]

bunkered for 2 month's I'd like to know how he did that

 

[StefanR5R]

Well, I just looked --- task deadline at SETI@home is a little under 2 months (53 days), so that was about the period for which he bunkered. The user didn't have top-of-the-line GPUs, but still had to store magnitudes more tasks than SETI@home's quota on tasks in progress allows. It's a per-host quota (times GPUs per host), hence the user set up multiple logical hosts, a.k.a. multiple client instances. I don't know if the user did all this manually, or scripted it; but the latter is definitely doable.

 

[Orange Kid]

Day two of Asteroids@home
Stretching our lead in the second spot. I don't know how we are doing it but it's great.
Tomorrow may be an interesting day to see who has any bunkers or even tasks left.


1 Gridcoin ........................……..25 ............21,721,440
2 TeAm AnandTech ..................18 ............10,154,400
3 SETI.USA ...........................….15 .........….7,782,240
4 Czech National Team .........….12 .............3,099,840
5 UK BOINC Team .....................10 .........….3,062,880
6 Rechenkraft.net .....................….8 .........….1,654,560
7 SETI.Germany .....................…...6 .............1,482,720
8 Planet 3DNow! ...........................4 .........….1,322,880
9 L'Alliance Francophone .........….2 ............….995,040
10 SETIKAH@KOREA .................1 .................778,080

 

[StefanR5R]

The third 18 hours:

Server status: of course ≈0 tasks ready to send, 160,000 tasks in progress, falling by 50,000 per day.

 

[Ken g6]

A fairly boring day for us, but check out the UK vs. Czech battle!

 

[Kiska]

Last year, @Kiska observed a rate of 6 tasks generated per second.

I believe this generation rate is gonna be lower due to the "lack of space" issue