Building decoupled polyglot applications with Vert.x
@tcrawley
Clojure/Conj 2013
What is Vert.x?
“Vert.x is a lightweight, high performance application platform for the JVM that's designed for modern mobile, web, and enterprise applications.”
What is a human?
“A Human is a bipedal mammal that procreates and turns food into waste."
What problems does Vert.x solve?
The JVM networking packages are byzantine
Efficiency requires concurrency, which is error-prone
Everything can't be asynchronous
Plumbing decoupled components can be painful
How does Vert.x solve these problems?
The JVM networking packages are byzantine
An abstraction over Netty
Efficiency requires concurrency, which is error-prone
Thread confinement
Everything can't be asynchronous
Sync/async hybrid
Plumbing decoupled components can be painful
Built-in event bus
Polyglot
- Java
- JRuby
- Jython
- JavaScript/CoffeeScript (Rhino, DynJS*)
- Groovy
- Scala*
- PHP*
- Clojure*
(* in-progress)
Why Async?
Why Async?
Reactor Loop(s)
Defaults to one/core
Verticle tied to one loop
Thread confinement
Distributed Event Bus
- Simple (String, int, byte[], etc) & structured (JSON) messages
- Publish/Subscribe
- Point-to-point
- Request/Response
- Clusterable (Hazelcast, browser)
- Transient
Sample App
This is where the demo happens
See https://github.com/tobias/vertx-clojure-conj-2013/tree/master/rivulet
Sample Application
Sample Application
Features
- HTTP(S), TCP, UDP, WebSocket servers
- Ditto clients, plus DNS
- Distributed Event Bus
- Timers
- Async FS
- Embeddable
Gains
- Hides Netty complexity
- Efficient
- Distributed Event Bus
- Thread confinement
- Polyglot
- Clusterable
Losses
- Hides Netty complexity
- Language ecosystems
- "Callback Hell"
Don't block…
Worker Verticles
Runs on a thread pool, so can block.
Still thread-confined by default.
Ancillary Projects
- Ring Vert.x Adapter - https://github.com/stream1984/ring-vertx-adapter/
- NoDyn - http://nodyn.io
- Yoke - http://pmlopes.github.io/yoke/
Resources
https://github.com/vert-x/mod-lang-clojure/
