Scala the Cloud Native Way...
summary
Linkerd is a Scala/Finagle/Netty service mesh used in production at Salesforce, CreditKarma and FOX that struggles with per-sidecar memory. Engineers switch the JVM from HotSpot to OpenJ9 and drop a typical per-host footprint from ~400MB to ~213MB and further to ~170MB with tuning flags. They explore ahead‑of‑time compilation with GraalVM (blocked by dynamic plugins) and rewrite the proxy in Rust to get sub‑millisecond latency. Core tools highlighted include Finagle, Netty, JVM tuning and a move to a per‑host deployment model.
AI-written summary · about
chapters (11)
- 0:00Start
- 2:45Why you would use it?
- 5:52The world is going cloud native
- 6:37Linkerd runs as a distributed proxy
- 7:41Alternative deployment: Per Host
- 9:16Ways to reduce JVM memory footprint
- 10:19Tuning JVM, Finagle, Netty • JVM flags
- 13:15Other OpenJ9 optimizations
- 16:31Don't use the JVM
- 18:12Current state of Linkerd
- 19:33Recap
related talks
06mInterview
47mScala: The Unpredicted Lingua Franca for Data Science
48mMassively Parallel Distributed Scala Compilation And You
30mDistributing systems in Scala with Fiangle
24mHigh Performance Serverless Functions in Scala
20mTop new CNCF projects to look out for
30mMigrating a monolith to Cloud-Native and the stumbling blocks
31mWhy the future of the cloud will be built on Rust
