9th Annual BOINC Pentathlon 2018 Strategy & Preparation

I bunkered ALL the projects that had a wingman before the project was officially picked. If only I had the hardware to complete them all (the ones picked) in time though. The others I had to abort.
 
This is interesting. We've been outperforming TC by about 200K+ per hour for the last few hours.

If they've just got a network connection down or they're bunkering to jump over P3D, then we're stuffed, but if they've moved to Asteroids to fend off SG then this could be close.

The most likely explanation is the former but I've fired up another four 72 vcpu EC2 instances just in case. I fear that, without yet more output, we'll still fall short but what the hell.

I'm going to hold on to the bunker until after it's too late for them to crunch any of the shorter Decic Field tasks. This is risky given that server issues could prevent a late upload but I feel releasing now would be pointless whilst they still have time to react.
 
if our production stays at 250k in the 5hrs or less we have left we are only going to gain around 800-900k points over their production. so who knows with your bunker and maybe if everyone was on NF :) but your right it could be close :) crossfingers
Either way I think we all did awesome :) Good Job Team :)
 
All this asteroids crunching has me close to 1 million points. Will end around 830,000. Maybe I should crunch for a few more days on it.:turtle:
 
I think i have several days of cleanup as i am not going to dump the tasks i don't need. I only dumped tasks this race if i knew that by the time i could come back to it it would be past deadline. I did set everything to no more tasks. Gotta get it cleaned up and ready for the FB sprint coming up ! Really enjoyed my first Pentathlon though.
 
Just so you know aborting tasks doesn't mean they go away for ever. They just get put back in the pool and sent out to someone else when they request work.

As for me. I aborted all the tasks I had and shut down all my servers and closed the manager on my other rigs that stay running 24/7.
 
Yeah I seen that too. Pretty sure it's not the first time he's been banned. You should probably say something to him in Discord.
 
It's not even just DC'ing..it is completely dissing the fact that there are guides for "shitcoins" being made. How dare unpopular coins be discussed here...
 
Well, I don't know about anybody else but I feel DC exhausted. It was good fun though and I think we all deserve damn good pat on the back for our best ever placement.

If we want to do better next year we ought to conduct a review on our strategy whilst it's fresh in our memory. My two penneth worth includes:
  • We need to get better at pre event bunkering. Skillz approach to pre announcement downloading of projects requiring a wingman sounds pretty good. I think I was fairly quick off the mark but not as quick as that.
  • Given the teams we're up against. we're not going to earn a podium place if we go for every event. I'd suggest dropping the marathon except for the initial bunker and those wishing to support without having to switch. It earns the same points as the other events but takes a lot more computing effort. We probably would have placed 10th or above in the Marathon anyway and would have had a much improved shot at a medal in the other cpu events.
  • I should have set up vms/multiple boinc clients much earlier but my daughters birthday events intervened and I ended up dual booting instead. Having two or more bunkers full of early downloaded tasks really helps on wingman projects.
End of post mortem/ramble.
 
Yes, I think getting as many work units as you can BEFORE the project is announced is key with the ones that require a wingman. I should have grabbed a LOT more and held onto them. The problem here is that most of these projects do not have a long deadline. What you could do is each day if the project isn't announced just abort the project(s) that are close enough to a deadline where you can't finish them all and grab new work. Rinse and repeat until the project is announced or you know its not going to be announced.

I agree on the marathon. If we didn't put so much effort into just that project I think we could have fared better overall. Case in point: The last three days. When we seen Team China ramping up like crazy on our six we should have just abandoned all together. The team behind them wasn't going to catch us and we could have gained 2 - 3 positions in Asteroids possibly with a top 5 finish in the end.

Make sure you have your systems prepped and ready BEFORE the event starts. Tweaking things and not running them during the event hurts. Some things are unavoidable though and that's understandable. For example. My layout of systems had been running fine since December/January, but when I loaded them all up with work when the event started I had a few power failures from tripping breakers. Something that never happened before in the past. Coughed it up as a fluke and it happened two more times before I finally moved a 4P box to a different circuit. I'm guessing the warmer temperatures had something to do with it tripping breakers that were otherwise fine during the colder months. That hurt the team a good bit during one of those outages as I wasn't home to catch it. No idea how long it had been down and I was away from the house for 9 or 10 hours.

Recruiting. I made posts in multiple forums asking for assistance. Even if you guys don't make the initial post, post in them the teams update, positions, etc... get it bumped to the top and make it seem like a big deal with lot of replies from different people. Everyone wants to get in on something that a lot of people are doing. I also send PMs to everyone I've recruited before in the past asking them to help out. In most cases they come. My recruiting also helps that I am active in other parts of the forums. Since General Mayhem is where most of my recruits come from is most likely linked to the fact that I am most active in that forum. More so than I am here in the DC sub forum.

I need to learn to setup VMs on Linux so I can get a LOT more tasks to run at once. Bunkering, even if it's not a big bunker it is still a good thing to do all the time. It doesn't show our true potential and other teams wont put as much effort into something they think they got. Teams did it to us this year. We need to do it right back.
 
Last edited by a moderator:
Here is my personal observation.

What's common between Rosetta and NumberFields? In both we came at #4, very close to win a medal and both projects have quorum 1. This is what we do best at that time. Pure brute force crunching power. Maybe there is a chance to get a medal if we ditch Universe and Asteroid early but would impact overall team position. Lots of frustration about pending validation work. I like project with quorum of 1 as 100% of effort put in is being converted to 100% pure result. With quorum of 2, if you put 100% effort, you are lucky to get 80-90% plus return. +1 with Skillz's comment on early download of tasks prior to project announcement to mitigate that frustration.

At some point in the race, we need to decide if
(a) we want to win a medal or two
(b) sacrifice overall team position.

Both are hard decisions and we as a team need to find a common consensus. We may choose the wrong path but that's ok as long as there is consensus and decision based on the situation/information at that point of time. That's call good teamwork and good communication.

SUSA has help from big boys from OC UK and XS. If you minus off contribution from these two teams, I would think P3DN would take the top spot. Good for SUSA to be able to attract such "talents" or does anyone know if they made a pact with OC UK and XS early prior to the race? If I may suggest this: what about forming a pack with TAAT and/or OCN for next year Pentathlon? Both have good people. Need to iron out finer details but just for fun, I would like to hear opinions on this. As usual the devils are always in the details.
 
Those that were running wingman projects with no bunker did pretty good. I think we were close to 100% right at guessing the projects early and alot of us could have been running prime / asteroids on gpu etc WAY early to "pre load", and then bunkering when announced. Couple days prior to marathon / other events being announced (before contest projects announced at all) start ramping up possible bunkers. <-- i realize this was somewhat suggested during.
Skillz is right about recruiting being a team thing. Myself and a few others were jumping in to post / like and it made a big difference in thread visibility but still sometimes i saw it go to page 2 in some forums and yeah its more genuine if more people are engaging in it. Recruit outside the forum / people who don't DC normally. I got 4 people from work to help and got over 500k points for marathon so spam your friends. Gotta say i learned a lot by doing / making mistakes and will do better next time on task / project management.
Also i still suggest we have more of a formal consensus for some things or at least have some kind of chart of decisions of direction to make it easier for those that don't read the whole thread(s)
 
and yeah we came out blazing on marathon but switched late and it cost us big later on. I think alot of that was because of indecision about going all out for marathon early on VS placing higher in all projects
 
I don't like the idea of hiding systems. Yes I know in these events people do it and I know some do it year round. To me, hiding your systems is like saying you "have something to actually hide". If they see I'm bringing a dozen PC's so what. If they want to bitch because I use an AWS server, so what. Let them bitch. Some people feel the competition will go out and pick up more rigs. I personally think they will do that anyways.... I don't see the real point in hiding them.
 
Jeeper final report, part 1.

Notable quotes from this report from an outsider perspective. I don't want to say if he is right or wrong other that we had this discussion about defending our 3rd spot in marathon.

[H] were left with #8 and at that with a great margin. Had they been prepared to sacrificed this, it might have been enough to reach Bronze at the Marathon.
Click to expand...

In the end, [H] were left with only #4. It looks as if they felt secure far too soon and invested too many resources in the City Run. This shows us once again why tactics can be so decisive in the Pentathlon.
Click to expand...
 
Those point totals that were dropped in bunkers for the marathon were very large at the end. Both from our team and Team China. I feel like to counter Team China we would have had to have all CPUs on the Marathon for the duration. And got lucky on Universe and Rosetta.

Edit: Can these same projects be picked next year?
 
Please see the rules on how a project is being selected here and the current tally on each project.

The most popular project will be the sprint event, i.e. the shortest event.
 
Pretty strong chance that WCG, YoYo, LHC, and Einstein will be there next year....lol

Those were 4 of the 5 last year and then two years before as well....
 
You must log in or register to reply here.
Back
Top