Concurrency + Distribution = Scalability + Availability(...)
summary
Claims that Erlang’s concurrency and distribution enable scalability and availability but require concrete architecture choices: fully connected distributed Erlang hits practical limits around 70–200 nodes, consistent‑hash rings (32–64 partitions) and Riak‑style sharding extend that, and Splunk‑style socket tricks push inter‑node throughput to ~60,000 req/s. Covers real tools (Erlang/OTP, distributed Erlang, RabbitMQ, TCP/SSL, AMQP), concrete node types (front‑end nodes, logic nodes, service nodes), and state strategies (share‑nothing, share‑some, share‑everything), with tradeoffs and retry models (at‑most/at‑least/exactly‑once) and production examples that succeed.
AI-written summary · about
chapters (8)
- 0:00Start
- 0:31Design Patterns
- 7:46Node Types
- 10:04Distributed Patterns
- 18:49Splunk
- 19:26Note Interfaces
- 27:36Retry Strategies
- 33:20Primary Primary Replication
