Maintaining highly distributed services with Erlang: 20y experiences in fintech
summary
Erlang powers a 20‑year fintech core that runs thousands of transactions per second, does daily zero‑downtime deployments via hot code upgrades, and survives regional, DB and cloud migrations (moved to AWS and switched to Postgres without downtime). Architecture uses supervisor trees and the let it crash philosophy; systems run 50–100 MB services up to 100 GB (historically 1–2 TB), process counts in the hundreds of thousands, and an event pipeline with 30 topics 400 partitions processing 1–2k events/sec.
AI-written summary · about
chapters (10)
- 0:00Erlang in fintech
- 1:24Erlang for distributed systems
- 2:50Erlang core concepts
- 8:12Erlang scalability model
- 13:47Fault tolerance in Erlang
- 15:33Supervisor trees
- 21:14Hot code upgrades
- 27:56Massive concurrency
- 30:22CL’s Cred system
- 35:57Summary
related talks
42mErlang Micro-Services with all the Buzzwords
37mFrom the ground to the stars
01h 00m10 years of microservices at FINN.no - and we still haven’t slain that dragon!
47mShipping a Replacement Architecture in Elixir
47mApplying “web scale” patterns (...)
54m$10K per Minute of Downtime: Architecture, Queues, Streaming and Fintech
45mMicro Services and Events: Friends or Foes?
35mFrom Monolith to State-of-the-Art Banking