Erlang: Bult for concurrent, distributed systems
1 like1,124 views
Erlang is a programming language and runtime system designed for building massively scalable soft real-time systems with requirements for high availability. It uses lightweight processes and message passing for concurrency and distribution. Processes have their own memory and mailboxes for communication, allowing fault tolerance through redundancy. Erlang is well-suited for building distributed, concurrent applications like chat servers and routing meshes.
Erlang: Bult for concurrent, distributed systems
- 1. Erlang Built for concurrent, distributed systems July 27, 2011 Ken Pratt http://kenpratt.net/ Wednesday, July 27, 2011
- 2. The philosophy "Erlang was designed for writing concurrent programs that "run forever"" - Joe Armstrong Wednesday, July 27, 2011
- 3. The philosophy Simple language Ultra-lightweight concurrency No shared state Local & remote communication via messaging Wednesday, July 27, 2011
- 4. The language Functional Pattern matching No variables! (well, they don’t vary, anyway) Compiled to bytecode, runs on VM Wednesday, July 27, 2011
- 5. The VM Processes Pre-emptive scheduler Memory manager Mesh networking Wednesday, July 27, 2011
- 6. Erlang == distributed OS? Erlang is not so much a programming language as it is a distributed Operating System. It’s like an ant colony with thousands of ants performing all sorts of different tasks. Erlang is to threads what garbage collection is to manual memory management -- it brings concurrent programming into the 21st century. Wednesday, July 27, 2011
- 7. Local architecture = light-weight process with a built-in mailbox Wednesday, July 27, 2011
- 8. Distributed architecture = light-weight process with a built-in mailbox Wednesday, July 27, 2011
- 9. Interesting architectures Having thousands of lightweight “threads” really opens up the architecture options. The main example application on the “Erlang and OTP in Action” book is a cache using one process per cache entry! Wednesday, July 27, 2011
- 10. Standard library Common patterns do all the concurrency heavy lifting for you (gen_server, rpc) Supervision trees Hot code deployment Wednesday, July 27, 2011
- 11. In production Chat backend SimpleDB Orchestration of Ruby services Routing mesh, logging system Wednesday, July 27, 2011
- 12. Erlang + Neo4j Erlang has graphs: http://www.erlang.org/doc/man/ digraph.html BUT, it’s no Graph DB Wednesday, July 27, 2011
- 13. Erlang + Neo4j Nerlo: http://github.com/nerlo/nerlo (Jinterface-based) Cali: http://github.com/dmitriid/cali (Jinterface-based) Both look pretty dead-in-the-water (no commits on either project in ~1 yr) BUT, Jinterface works well, so does Mule ESB. So in the worst case you could write your Neo4j adapters in the JVM and bridge to Erlang with a biz-logic-level API. Wednesday, July 27, 2011
- 14. Resources Wednesday, July 27, 2011