Hacker News new | ask | show | jobs
by paulryanrogers 991 days ago
What are the tradeoffs? Guessing based on the stack it's eventual consistency
1 comments

No. It's a distributed database. Transactions are committed in memory, and then (by default every second) committed as a group to disk. This gives you phenomenal write throughput and low latency, but the tradeoff is that if a whole cluster goes lights out, you could lose a second of data. It's kind of like turning off fsync for InnoDB.