Hacker News Reader

Very innovative idea. Not sure if it's in the article, but did you bring CockroachDB down? Also, for anybody trying to repeat that experiment, you should have a look at Jepsen (https://jepsen.io/analyses/cockroachdb-beta-20160829)

View on HN

We definitely could have. That said, to make it a "fair fight", where the cluster was trying as hard as possible to survive, we would have had to mess around with the amount of data stored on it. The current implementation works with a fixed amount of data, but we toyed with idea of starting with some large amount, which means nodes would take longer to replicate, and then decreasing it as we went, until roaches spawned so fast they overwhelmed you (but ultimately didn't implement it because of time constraints)

Definitely going to take a look at that!

Our foundation with respect to distributed systems isn't really formally rigid (I'm still trying to understand how Raft and Paxos work) and looking at this will probably help a lot!

I would suggest https://raft.github.io/

That has a nice, interactive visualization along with links to the paper itself, implementations, and talks about raft.

Reply on news.ycombinator.com