Sprints in 2019

Locked
User avatar
sebastien
Steering committee
Posts: 119
Joined: Thu Jan 01, 2009 2:59 pm

Sprints in 2019

Post by sebastien »

This year, some project servers crashes during sprint. I opened this topics to find solutions to this problem for 2019.
Some users suggested some solutions in other topics:

- Alert projets administrators before sprint. (Possible if some users accept to do this job)

- Exclude small projects admissible to sprints. (contrary to the purpose of Formula BOINC)

Please discuss here about this proposals and / or other proposals.
Image
JeromeC
Posts: 3
Joined: Sun Dec 28, 2014 5:26 pm

Re: Sprints in 2019

Post by JeromeC »

All teams are supposed to be aware about the sprint at the same time, a reduced number of days before it happens (how many ?), would this time be enough to inform the project admins ? (since some members of those teams would be the ones that inform them, they can't be informed themselves before the official announcement is done)

And what happens if an admin then answers "I don't want to be the sprint project" ?


Excluding arbitrarily some projects would be a problem : what is the limit of a "small project" ? and what if some of them feel they can handle a sprint anyway ? I agree with you when you say this is contrary to the purpose of FB...
Image
User avatar
Coleslaw
Posts: 48
Joined: Sun Oct 21, 2012 5:02 am

Re: Sprints in 2019

Post by Coleslaw »

- Exclude small projects admissible to sprints. (contrary to the purpose of Formula BOINC)
I see having a sprint that brings the small projects down to be even more against the core purpose of Formula BOINC. Marathons can still support the little guy without doing the harm that the sprints cause. I think history is a pretty good indicator of what projects have not been able to support sprints. So, asking what does "too small" really mean isn't really a fair question. Some common sense can be used here. Quite frankly, I don't want to see any project less than a year old as a sprint because I want them to focus on stability and growth first. There is just too much of a learning curve when starting out to be fully aware of what it entails to handle a sprint.

I also don't see how hard it is for someone within FB to take 3 minutes of their time to post something in the projects forums or PM an admin about the event. Some of the very users that continue posting in these dead forums have already been trying to give admins warning because we are sick of failed sprints. The admins don't typically look at FB's calendar of events. From what I have seen and been told, the projects really haven't been asked for inclusion anyways. Some appreciate the Marathon opportunity but some do not care for the sprint behavior due to not having warning to prepare for the need of additional resources. You end up causing these projects to waste man hours and possibly even work done due to down time by not giving them proper warning.

If we are "fixing" the sprints, why not go ahead and address the other concerns that have been brought up for years. Like finding a way to prevent a project that ends at the beginning of the year or goes dormant for several months from being worth the same points as a project that is active year round? Squatting on dead project points may be a tactic but I can tell you it is a big complaint from teams that actually want to focus on FB. The teams that use scripts to hammer (DDoS) project servers to get the very few work units get the reward. Are we saying FB supports this way of play? Because that is what teams are having to resort to in order to win. These scripts are also many of times used in combination with Virtual Machines that can be set to only focus on those projects and be combo'd with NCI projects to be put to better use. I know you had concerns at WUProp about some of the various misuses.... I know I have and still do run some VM's that their sole purpose is to catch work from hard to get projects and sometimes I have to resort to all sorts of trickery or configurations to get it. I cannot see how this is "good" for the projects.

I do not see a good way to prevent the team hopping outside of people just agreeing to not be so shady in their tactics. This tactic was actually used long before [H]ard|OCP used it. People need to understand that sometimes people will be members of more than one team. So, they have split allegiances. Sometimes teams have multiple team names at a project like we do. So, one of the team names may show up under the team stats across the board that doesn't belong. Yet you will still see the correct team's stats reflected in individual users stats at the stats sites. You cannot use the stats sites correctly with how the sprint stats are handled here. They just aren't necessarily going to be accurate.

The concept of 3 Leagues is nice but takes too long to correct itself as far as teams and their capabilities. To be honest, [H]ard|OCP and a few other teams should have been put in League 1 2 years ago. However, when the decision was made was probably at a low interest time and so didn't fairly lump teams where they truly fit. We have spent 2 years feeling like we have had virtually zero competition here. Yet we know there are teams that can compete such as XtremeSystems. They just don't care enough to put forth the full effort. That and they support other teams as well. So, next year we will finally be in League 1 where we should have been and will be able to compete with the few actual active FB teams that there is. I fear that after the last 2 years of bad sprint experiences and non-competition may be wearing thin. We built a lot of momentum this year but have had some second thoughts about continuing. I would suggest in future years to make it a sign up requirement so that only teams that want to compete are listed (yes each year).

The only real problem I could see coming up with notifying projects in advance (as noted above) is if they happen to either have their own team or are a part of a team. That would make some question their judgement even if it isn't really warranted.

Some positives to include are: I really like the idea of sprints being tied to Formula racing as it gives teams a more balanced approach to start times. People willing to run work during warmer months will have a slight advantage as well.

The sprints are great practice for much larger challenges and keep you learning tactics for several projects rather than same ones all the time. This also encourages people to learn better ways to utilize their hardware to be more efficient and squeeze more science from it. You can learn a lot from a 3 day challenge on the behavior of a project and how they respond to pressure. Things like multiple clients, config files, VM's, Hyper Threading on/off, project hard limits like number of work units a client/account is allowed per day or at any given time, validation turn arounds and likelihood of having a wingman finish within 3 days (4), etc...

I like the suggestion above about possibly using sub projects. However, this does bring other concerns like giving too much focus on certain projects which can be seen as unfair. Also, how do you choose the sub projects and what happens if the sub project ends or doesn't have enough work to support? Yes, I know the common sense answer would be to make adjustments as these things come up, but we really don't see any adjustments during the year right now. I personally would be ok with a sub project change as I'm not dead set on running any one specific science. Others may be. I honestly haven't looked into which projects break down sub project stats for everyone, but I'm open to the idea of using it.

Since only project admins have enough control to really refine restricting things, I will try and avoid making any suggestions for mimicking what individual projects like PrimeGrid do for their challenges.
User avatar
bcavnaugh
Posts: 112
Joined: Fri Sep 29, 2017 12:06 am

Re: Sprints in 2019

Post by bcavnaugh »

Coleslaw wrote: Mon Oct 08, 2018 4:00 am To be honest, [H]ard|OCP and a few other teams should have been put in League 1 2 years ago.
Only because of the of GPU Miners you pay :lol:

Or teams were in League 1 3 years ago
http://formula-boinc.org/index.py?year=2015
Gridcoin Put it all out of whack
Image
Crunching@EVGA The Number One Team in the BOINC Community. Folding@EVGA The Number One Team in the Folding@Home Community.
Ken_g6
Posts: 10
Joined: Tue Jun 26, 2018 6:34 pm

Re: Sprints in 2019

Post by Ken_g6 »

I suggest you take a page from the BOINC Pentathlon. This year they asked all the projects, well before the start of any challenge, whether the project admins wanted to participate. I suggest you ask in December, before the schedule even appears on the site.

Now, yes means yes and no means no. BOINC Pentathlon decided that a null response also meant no. You can decide for yourself what to make of a null response. On one hand, if the admins don't care enough to answer yes, why should we care enough to participate in their project? On the other hand, you might not get enough "yes" responses to come close to filling the sprint list. BOINC Pentathlon barely got enough to fill out their list. You might have to consider running the same projects two or three times in a year.
User avatar
bcavnaugh
Posts: 112
Joined: Fri Sep 29, 2017 12:06 am

Re: Sprints in 2019

Post by bcavnaugh »

Skillz wrote: Mon Oct 08, 2018 8:37 pm
bcavnaugh wrote: Mon Oct 08, 2018 8:22 pm
Coleslaw wrote: Mon Oct 08, 2018 4:00 am To be honest, [H]ard|OCP and a few other teams should have been put in League 1 2 years ago.
Only because of the of GPU Miners you pay :lol:

Or teams were in League 1 3 years ago
http://formula-boinc.org/index.py?year=2015
Gridcoin Put it all out of whack
No one is paying anyone. Where did that even come from? He paid for his hardware and he pays for his electricity. He can do whatever he wants to do with them.
Did you miss this :lol:
Image
Crunching@EVGA The Number One Team in the BOINC Community. Folding@EVGA The Number One Team in the Folding@Home Community.
User avatar
bcavnaugh
Posts: 112
Joined: Fri Sep 29, 2017 12:06 am

Re: Sprints in 2019

Post by bcavnaugh »

Ken_g6 wrote: Mon Oct 08, 2018 9:28 pm I suggest you take a page from the BOINC Pentathlon. This year they asked all the projects, well before the start of any challenge, whether the project admins wanted to participate. I suggest you ask in December, before the schedule even appears on the site.

Now, yes means yes and no means no. BOINC Pentathlon decided that a null response also meant no. You can decide for yourself what to make of a null response. On one hand, if the admins don't care enough to answer yes, why should we care enough to participate in their project? On the other hand, you might not get enough "yes" responses to come close to filling the sprint list. BOINC Pentathlon barely got enough to fill out their list. You might have to consider running the same projects two or three times in a year.
Or Reduce the Number to 12 Sprints and have 1 Sprint a Month.
Image
Crunching@EVGA The Number One Team in the BOINC Community. Folding@EVGA The Number One Team in the Folding@Home Community.
Ken_g6
Posts: 10
Joined: Tue Jun 26, 2018 6:34 pm

Re: Sprints in 2019

Post by Ken_g6 »

A teammate of mine suggests a hybrid approach:
xii5ku wrote:Regarding prior notification of project admins, here is an example of a simple routine:
A few weeks before a sprint, select three projects randomly. This can be done on a relaxed schedule, but not too late. (Maybe a bit more than a "few" weeks are advisable.)
Send an e-mail to the administrators of each of the three projects, asking whether it would be convenient to hold a contest at the given time frame. (Also tell the admins how much contributor capacity was seen at recent sprints, and that their project is one out of three taken into consideration this time.)
As soon as the first positive reply is received, select this project, inform the admin of the choice and thank in advance, and send a friendly cancellation note to the other two admins. (These other two choices could be elected again for a subsequent sprint.)
If only negative replies are received, cancel the sprint.
If no replies are received at all, either cancel the sprint, or select one of the projects anyway, and state in the sprint announcement that no contact to the project admin could be established prior to the sprint.
(copied from the Asteroids forum, slightly edited)

--------
If the idea gets accepted that sprints could be canceled when there is no response from admins, or at least in case of negative responses, I suspect quite a few sprints in a year could end up canceled. Which would be disappointing to prospective sprinters. But sprint projects without work, too slow work generators, a defunct validator, an acoustic coupler for internet connection, ... are a bit of a disappointment too.
Orange Kid
Posts: 8
Joined: Thu Apr 13, 2017 2:37 am

Re: Sprints in 2019

Post by Orange Kid »

Simple fix: eliminate the 24hr announcement.
Take the initial stats reading, announce the project and BAM!, away we go. Just don't miss the announcement as has happened in the past.
This will eliminate the bunkering and the first day huge dumps.
That said:
Nobody is forcing anyone to compete, If you don't want to do it, don't. Some of us enjoy the competition and spend GREAT amounts of money not only on hardware, but also the costs of running said machines. To allow someone else to dictate to us what we can do with our resources is ludicrous. If we want to split our effort between different teams it is our right, not yours to tell us that we cannot.
That's the end of my rant.
User avatar
ChristianVirtual
Posts: 47
Joined: Tue Aug 08, 2017 2:58 pm

Re: Sprints in 2019

Post by ChristianVirtual »

No heads up is not fair for those in timezones not allowing to activate the proper settings. E.g. sleeping or working when a sprint start. Then we need to also have a rolling start of sprints that each time zone gets it for feasable Times.
Any direct message might be subject of publication; sender's consent assumed
Locked