Hello, in this article I wrote my participation report for the event. We didn’t make a breakthrough, but I hope this post will motivate some of you to participate the next time.
Cheers!
-Tristan
Hello, in this article I wrote my participation report for the event. We didn’t make a breakthrough, but I hope this post will motivate some of you to participate the next time.
Cheers!
-Tristan
Setup an infra: we could have used a VPN to share compute cores (though we only had two i5 laptops and one ryzen desktop), and to setup a database to keep track of the solutions.
This should be trivial. Tailscale free tier offers up to 100 devices in the net, this should be enough for prototyping.
What is not trivial, is cat-herding all of them into doing work.
I read a lot of reports that said things like “we’ve left our brute-forcer running went to sleep while it churns better solutions”. Some teams are even reserving time on their workplace/uni clusters.
With a mining winning pool large enough it should be possible to rock the lightning round hard. And the actual VPN may not be necessary if you distribute some auth tokens for a supervisor service.
The challenge here is to keep solutions versioned and worker nodes updated.
Make a bigger team: that way we could have assigned the problems and tried different strategies.
I’ve completely missed the event this year. It would be a good idea to bat-signal the community next time. Maybe someone can bring their big iron and/or big brain in time.
Thank you for the suggestions. Without knowing what the next contest will be, we may already define some organizational techniques.
For example, with a team of n member, we could dispatch the problems using mod problemNumber n == 0
to find the initial strategies.
Then I was thinking about having a database to store the solutions so that we could easily try running the strategies on the other problems. It was quite painful to use git for that, and having some query filter would have been great. Then we could use a dedicated process in charge of submitting the solutions, perhaps based on the global scoreboard…
It’s hard to tell in advance what will work the best, and it might be necessary to closely watch the solving process, then the limiting factor is brain cycles
If the time allows I’d like to try again, then I’ll send a call for participation, I hope we’ll do better next time!