Tolerating partitions

@captheorem

Wrong answers only

Joined January 2019

6Posts 2Followers 1Following

Tolerating partitions

@captheorem

Jan 12, 2021

Do you remember when you joined Twitter? I do! #MyTwitterAnniversary

Andrew Clay Shafer 雷启理

@littleidea

Sep 27, 2020

eventual consistency means people will start getting all the great CAP theorem jokes

I say it's probably fine: - each object is single-threaded - each object exists in one place at a time Minimal coordination demands. But this sounds like it could potentially be a headache: - Durable Objects will combat this problem by auto-migrating

ahmetb

@ahmetb

Sep 30, 2020

Cloudflare Workers Durable Objects opened a can of worms for me: blog.cloudflare.com/introducing-wo… It's a storage for the edge workers with: * low-latency * strong consistency * not zonal, but global replication. But CAP theorem says... 🤔 🤔 🤔

Tolerating partitions reposted

Abstract Cloud

@eric_brewer

Oct 2, 2020

I have been asked this a great deal. I think the real issue is how *many* services you have, not their size. If you have a few, manage them how you like. When you have 100 or 1000 (or 1M like Google), you need infrastructure help, especially if you also want release velocity.