Building Scalable Web Services with Vert.x

Web services são aplicações que fazem uso intensivo de I/O. Tradicionalmente tais operações de I/O são implementadas de forma síncrona e a única forma de conseguir paralelismo neste cenário é a utilização de múltiplas threads do processador. Esta abordagem consome uma grande quantidade de recursos e não escala adequadamente.
Vert.x é um toolkit para JVM que permite a utilização de I/O assíncrono. Desta maneira, é possível construir aplicações de real-time altamente escaláveis e utilizando o mínimo de recursos.

Transcript

1. z Building Scalable Web Services with Vert.x Marcelo Pereira (Backend Developer at Pismo)
2. z Summary § Nature of web applications § I/O processing § Blocking vs non-blocking § Reactive programming § Vert.x
3. z The nature of web applications
4. z Map of the island of I/O-bound programs Source: Node.js The Right Way: Practival, Server-Side Javascript That Scales (Wilson, Jim R. 2013)
5. z I/O bound programs § I/O-Bound programs are constrainer by data access § Adding more processing power or RAM often makes little difference
6. z I/O processing
7. z I/O processing characteristics § Extremely slow compared to CPU processing § Depending on the device it can incorporate mechanical parts as hard drives § Even SSD makes use of long and slow buses
8. z Data transfer rates compared 600 25600 175000 0 20000 40000 60000 80000 100000 120000 140000 160000 180000 200000 SATA 6.0 Gb/s DDR4 RAM L3 Cache Velocidade em MB/s Velocidade em MB/s
9. z Blocking vs non-blocking
10. z Achieving high paralelism § Multi-threaded code § Single-threaded code and non-blocking I/O
11. z Pub Analogy
12. z Blocked taps gastropub
13. z Async beer lounge
14. z HTTP request processing example
15. z Apache/Nginx comparison Source: https://help.dreamhost.com/hc/en-us/articles/215945987-Web-server-performance-comparison
16. z Apache/Nginx comparison
17. z Reactive programming
18. z ReactiveX § Non-blocking APIs are event-driven by nature; § ReactiveX is a library for composing asynchronous and event- based programs by using observable sequences.
19. z Vert.x
20. z What is Vert.x § It is a tool-kit for building reactive applications on the JVM.
21. z Event loop
22. z Verticles
23. z Event bus
24. z Code example https://github.com/marceloandradep/vertx-example
25. z Challenges § Stacktraces are useless for debugging § Need of criterious logging with correlation Ids § Database deadlocks § Write, run and debug cycle doesn’t work § Tests, tests, tests, tests...
26. z Thanks! marcelo.pereira@pismo.io marceloandradep@gmail.com