Scalability concepts: read-after-write consistency

avikdas.com

3 pointsakdas4 comments

I've learned a ton working at scale, and I wanted to share some of that knowledge with others.

To be honest, this is the type of knowledge I actually had to unlearn when I decided to try my hand at early-stage startups. Unfortunately, this is also the type of knowledge that's helpful to go from a startup to a big tech company, where you get the chance to learn this in the first place! Hopefully my writing can help bootstrap that knowledge for others.

See Google’s Spanner Database.

Absolutely, it's something I researched while writing this article (since I knew someone would bring it up :)). Ultimately, I wanted to talk about concepts more than specific technologies, so I avoided going into detail on Spanner.

I don't claim to understand the underlying tech for Spanner, since I haven't read through the research paper fully. But my understanding is that write latencies are increased, which is a trade-off that has to be made. No matter what, _some_ trade-off is necessary, unfortunately.

Yes, "write latency" exists if the paxos group is down.

https://storage.googleapis.com/pub-tools-public-publication-...

Reply on news.ycombinator.com