Distributed Systems Without Magic: What CAP and PACELC Really Say (Category SystemDesign)

19 In March, I gave a lecture to Central University students on a wide range of topics – we talked about the CAP and PACELC theorems, we discussed how consistency in these theorems differs from consistency in ACID, we talked about how the declared guarantees of consistency are tested.( On the example of the project from Jepsen)Cassandra was discussed as an example of a distributed database with tunable consistency. Most of the theory was taken from articles from my website system-design.space, and after the lecture, the guys had a seminar where they could spin Cassandra with their hands and try the work of these theorems in practice.

I loved the story myself, but in this one. transcription I decided to focus not on theory, but on the simple idea that distributed systems don't have magic. I like Arthur C. Clarke’s Third Law.

Any sufficiently developed technology is indistinguishable from magic. But if you are a practitioner and an engineer, there is no database, no network topology, no architectural diagram that allows you to be always accessible, always consistent, and always fast at no cost. The architecture of a distributed system begins when we stop looking for magic and start honestly designing compromises. CAP, PACELC, consistency models, Jepsen and Cassandra are essentially different ways of saying the same reality.

#Architecture #Management #SystemDesign #Software #Engineering #Database