Formula BOINC Sprints 2021

Page 6 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
If my memory serves me right, this project has trouble handling sprints.
Your memory serves you right.
It looks like the work generator can't keep up with the demand for tasks.

Munin :: Munin-Node :: Munin-Node :: results tngrid
Yes, there is a well documented rate limit of the work generator. (Refer to the TN-Grid message board, and other forums including this one.)

If I was interested at all in how Formula Boinc is managed these days, just the bit of info that they are holding a TN-Grid contest would give me much of what one might want to know.
 

Icecold

Golden Member
Nov 15, 2004
1,090
1,008
146
I didn't set a backup project this morning when I switched over to TN-Grid and didn't have time to check on it during the day. I just received my very first 5 tasks 12 minutes ago with my 5950x having ran dry all day.
 

biodoc

Diamond Member
Dec 29, 2005
6,257
2,238
136
I'm finishing up my TN-Grid caches and switching to another project.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,480
14,434
136
Same here, though my cache is only 5 tasks :oops: I may switch back to it after it starts if it seems like the work generator is keeping up. It was mentioned on the FB forum, so Timbo is aware of the current lack of tasks at least, FWIW - https://formula-boinc.org/forum/viewtopic.php?p=2112#p2112
This is exactly why I don't want to participate in some of these. Why have a "sprint" when you can't even get tasks to compete ?
 

Icecold

Golden Member
Nov 15, 2004
1,090
1,008
146
I don't really disagree and it's been one of the main points of contention between Formula BOINC participants at least in recent history, but just playing devil's advocate here - there are not that many projects that can support a FB sprint, or the various pentathlon events. It leads to a situation where to avoid just running Primegrid, WCG, and a handful of others back to back they have to make decisions to run projects that may potentially have issues providing enough tasks for all participants. The pentathlon organizers are much more thorough in this than FB, but they're also not running a year long event with 20+ events like how FB is currently structured.

This is then compounded by the fact that when a project struggles to cope with the number of participants, savvy users that really want to win may edit their number of CPU's, or try to refresh BOINC a ton to get tasks before other users do, so it somewhat causes exponential issues.

There's really not great answers to any of these issues other than just being more selective about the projects chosen, but it certainly would be boring if it was just a rotation of the same 5 or so projects. It's also not necessarily very fun for a lot of people if it's not competitive. Just casually running a project without much care on winning or losing is not super exciting for a lot of people, so it leads to super competitive behavior which sometimes turns into anti competitive behavior like trying to grab all available tasks before other competitors can.

I really have no interest in having to work at it to try to get tasks, etc. though, so I'm going to switch off of it for now and check back in later and see where it stands, and if there are enough tasks I'll throw a couple machines on it at that time.
 

Fardringle

Diamond Member
Oct 23, 2000
9,184
753
126
I have to say that I'm not sure how TN-Grid got selected. We (Sprint Committee) decided some time ago that it is one of the projects previously listed as a potential sprint that really can't handle a sprint so it hasn't been a voting option. So either the selection algorithm messed up, or maybe Sebastien picked it on his own for some reason...
 

Icecold

Golden Member
Nov 15, 2004
1,090
1,008
146
I have to say that I'm not sure how TN-Grid got selected. We (Sprint Committee) decided some time ago that it is one of the projects previously listed as a potential sprint that really can't handle a sprint so it hasn't been a voting option. So either the selection algorithm messed up, or maybe Sebastien picked it on his own for some reason...
It shows 1 positive vote here and no negative votes - https://formula-boinc.org/admin/sprint.py?id=15

That doesn't contradict what you said at all, just providing any extra information I'm aware of. It does seem concerning that there is only 1 vote either way on each of those projects.
 

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
BTW, the sprint won't do harm to the TN-Grid project technically. As far as I recall, the server used to be able to handle lots of work requests. (It merely did not grant new work at most of the requests during competitions, due to the limited work generation rate.) That is, from the TN-Grid project's perspective, a competition leads to a situation in which all of the work which could be done, is done.

(Sounds ideal for the project. But there is also a potential "soft" negative effect for the project: Contestants as well as regular contributors could get disgruntled during such a contest and perhaps leave the project, which would of course be bad for the project once the contest is over.)

However, from the perspective of those who want to compete, a TN-Grid contest is nonsense. Or at least for those who want to compete normally by means of their donated computing capacity. To win a TN-Grid contest, this is what it takes: Set up multiple client instances and hammer the server with work requests by running scripts with each instance. Your computing capacity is secondary to that. Just use as many client instances as required to keep your computers busy all time.

There are probably people who enjoy contests of this kind, but I presume that these people are a minority among the target audience of DC competitions.
 
  • Like
Reactions: Ken g6 and biodoc

Icecold

Golden Member
Nov 15, 2004
1,090
1,008
146
I've been trying to run FB "casually" the last couple sprints. There are a few reasons why, but I mostly thought it might just be fun to throw a couple machines on it and see where our standings end up since we're not trying to really compete at FB and it's just me and 1 or 2 other people throwing a couple machines at it.

Honest feedback(and I will hopefully figure out some suggestions from this and post to the FB thread about it - I'm genuinely not trying to crap on FB but just formulate my thoughts) - it can be pretty demoralizing competing casually when you get a competition that there's no tasks. I've been sitting here clicking 'update' in BOINC manager to try to get tasks, and I'm getting nothing. I know how to script/batch file boinc to do update requests frequently, change NCPU, and I have a crap ton of PC's in my house, and I have extra BOINC instances waiting on BOINCTasks I could use to hammer the server to get more tasks and try to compete at a high level.. but I was just trying to compete casually, which I imagine is like 95% of people in BOINC competitions. If I'm some random person that isn't super familiar with BOINC but maybe has a newer higher end processor and wants to see how it does in a competition and I get no tasks because it's a project that barely outputs any tasks and they're all grabbed by people scripting it to take all the tasks.. I would imagine it turns me off of the whole thing. I refuse to hammer the server to try to grab tasks when I'm not even trying to win anything or compete at a high level, so instead I'm just sitting here clicking 'update' in the BOINC Mgr GUI and not getting a single task and feeling like the whole thing is kind of nonsense.

Edit - on the 100th update click I got 4 tasks on a 32 thread machine. .and I'm actually pretty happy about it :p

Edit 2 - having given it some thought, there's no productive suggestions other than not picking projects that cannot handle a sprint so I'm not going to waste anybody's time by posting that to the FB forum. I'm going to assume based on @Fardringle 's post that maybe the project selection algorithm messed up and Sebastien just didn't catch it in time, which if so isn't a huge deal. It does make for a boring 'sprint' when you can't even get tasks though

Edit 3 - this is absurd. I had 4 tasks on a 3950x, those completed and now I have 2 tasks.
 
Last edited:

cellarnoise

Senior member
Mar 22, 2017
709
394
136
You DC' ers, / bastards? Kidding, and love you all on this team.

A bit into my brain... I think of starting all I have on a challenge, which can be a bit of work as it is all by hand and self tuned...

No, I don't want to have to custom tune / write scripts to get work. Not Yet, but maybe I be full crazy over time?!!!

No, please no. I struggle to just keep my junk running and I might over time ease up on the voltage / frequency on cpu / memory / GPU eventually?

I supper stingy, but to write custom scripts to get work? That sounds like easy work, and I don't like easy :)(
 

Skillz

Senior member
Feb 14, 2014
911
929
136
I have to say that I'm not sure how TN-Grid got selected. We (Sprint Committee) decided some time ago that it is one of the projects previously listed as a potential sprint that really can't handle a sprint so it hasn't been a voting option. So either the selection algorithm messed up, or maybe Sebastien picked it on his own for some reason...

It's almost as if someone is purposely trying to destroy FB. I mean, this isn't the first time this year this has happened with a project that was known ahead of time that it can't handle and was even discussed as not being capable.

Ooh well.
 

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
I supper stingy, but to write custom scripts to get work? That sounds like easy work, and I don't like easy :)(
You are right in the regard that this is easy. Practically all teams who take part in DC competitions circulate primitive "update spam" scripts among them by now. That is, the difficulty of implementation is reduced to cut and paste.

By primitive I mean that these scripts simply force a project update in an infinite loop at a fixed period, without regard whether the client has work or does not have work. This can be written as a one-liner command line call.

(Edit: Obviously, a less primitive script would first check how much work the client currently has, and only issue a project update when there is not enough work queued. It is safe to assume that almost no team has implemented this small refinement in their scripts.)

The other part which I mentioned, using several client instances, involves a bit more work and insight than simple cut and paste, but how to do it is documented in many places. (In a TN-Grid contest, multiple client instances per computer would be used in order to partition a large and/or fast computer into several small and slow computers, because small and slow computers will have an easier time to maintain a full workqueue when the project's work generator is small and slow. Or put in other words, given the minimum waiting period which the project server enforces between work requests, using n client instances per computer effectively allows to divide this waiting period by n on this computer = issue n times as many work requests from this computer compared to a single instance.)

Edit: And of course none of this gets more work done for a project like TN-Grid, in the end.
 
Last edited:

Icecold

Golden Member
Nov 15, 2004
1,090
1,008
146
The sprint is Yoyo@home. If my parts come in tomorrow that I'm expecting I may have a machine to throw on it.
 

crashtech

Lifer
Jan 4, 2013
10,521
2,111
146
Yoyo is on my list of projects I want to do better in, but I'm still hesitant to give FB unearned legitimacy by participating.
 

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
The yoyo@home owner is on vacation. (Though maybe he still takes note of the contest and adjusts the server if necessary.)

Basic yoyo@home server specs were mentioned recently in the SiDock message board:
Dual core CPU, 8 GB RAM, spinning magnetic disks.

yoyo@home participated in FB sprints and in Pentathlons (2019, 2017, 2015…). I don't recall whether or not there were server performance issues — probably there were no serious ones —, but yoyo@home's means to maintain server performance include a low limit on tasks in progress per host, the collapsing of multiple client instances per system into the same host record (AFAIR this was introduced later; without it the former measure wouldn't be very effective), and extremely early deletion of completed workunits from the database.

Edit: Due to that, my personal opinion is that there shouldn't be contests held at yoyo@home if a participation at the level of FB 2018+ (let alone Pentathlons of the past few years) is expected or desired.
 
Last edited:

Fardringle

Diamond Member
Oct 23, 2000
9,184
753
126
I'm not certain since he hasn't posted ANYTHING since July 17th even though he has been logged in to the site regularly, but it seems like Sebastien is ignoring the Sprint Committee votes and just picking whatever projects haven't been run for a long time, even if there is a very good reason why they haven't been selected as a sprint project...
 
  • Wow
Reactions: Ken g6