Living in a Polyglot World
summary
A concrete claim: the server API is rewritten from Ruby Unicorn to Erlang Webmachine and cuts memory and latency dramatically. Measured numbers: Ruby workers sit ~100 MB idle and ~204 MB under load leading to ~19.2 GB total; Erlang workers sit ~19–75 MB and total ~600 MB. The wiggle in response time (median ~500 ms) disappears. Storage changes include a live CouchDB to MySQL migration and the full port takes ~a year with the live migration phase stretching from an expected 3 hours to ~3 days.
AI-written summary · about
chapters (15)
- 0:00Start
- 5:22Recipe
- 7:20Resource
- 9:58What is Chef!
- 12:50Client Design Goals
- 15:04Languages to consider
- 30:32CouchDB Uptime
- 37:54compaction
- 40:23What we need in a data store
- 43:25Live Migration in 3 Easy Steps
- 44:02Migration Tool
- 44:33Scripting with gen_fsm
- 45:33Database Load Average
- 46:27CouchDB Write Requests
- 46:39CouchDB Network Traffic
related talks
43mPolyglot Paralellism: A Case Study in Using Erlang and Ruby at Rackspace
42mWeb API's with ERLANG a Ruby Dev's POV
47mBERT and Ernie: Scaling your Ruby site with Erlang
48mBERT and Ernie: Scaling your Ruby site with Erlang
34mHow to scale a Ruby webservice
45mMigrating an Invoicing System to Elixir/Erlang
01h 03mRuby at Scale
28mRuby + Elixir: Polyglotting FTW!