The Reactive Landscape Clement Escoffier, Vert.x Core Developer, Red Hat

Transcription

1 The Reactive Landscape Clement Escoffier, Vert.x Core Developer, Red Hat

2 Reactive all the things??? Elasticity Manifesto Actor System Asynchrony Programming Events 2 Message Spreadsheets Resilience Scalability Back-Pressure Streams Data Flows extensions Asynchronous Observable Reactor Spring RX Java

3 3

4 FAKE! 4

5 FAKE! End of the flow of control? 5

6 FAKE! End of the flow of control? or just the comeback of asynchrony? 6

7 The reactive landscape Manifesto, Actor Messages, Resilience Elasticity, Scalability, Asynchronous Reactive Systems Akka, Vert.x 7 Reactive A software showing responses to stimuli Reactive Programming Streams, Flo ws, Events, Spreadsheet s Asynchrono us Reactor, Reactive Spring, RX, Vert.x Data Flow Back-Pressure Non-Blocking, Asynchronous Reactive Streams Akka Streams, RX v2, Reactor, Vert.x

8 The reactive landscape Manifesto, Actor Messages, Resilience Elasticity, Scalability, Asynchronous Reactive Systems Akka, Vert.x 8 Reactive A software showing responses to stimuli Reactive Programming Streams, Flo ws, Events, Spreadsheet s Asynchrono us Reactor, Reactive Spring, RX, Vert.x Data Flow Back-Pressure Non-Blocking, Asynchronous Reactive Streams Akka Streams, RX v2, Reactor, Vert.x

9 Mind the step Why my simple application is such a mess... My Application Some Software 9 Some Services

10 Modern software is distributed... Remote interactions everywhere My Application Some Software 10 Some Services

11 Distributed systems fail How to stay responsive in face of failures, under varying workload My Application Some Software 11 Some Services

12 Distributed systems fail How to stay responsive in face of failures, under varying workload My Application Some Software 12 Some Services

13 Reactive Systems => Responsive Systems

14 Reactive Manifesto 14

15 Asynchronous message passing Sends to an address 15 Message backbone Subscribes to the address

16 Asynchronous message passing tic as El Message backbone ity Not blocked while waiting for a reply Resilience 16

17 Reactive Systems from the trenches My Application Some Software 17 Some Services

18 Asynchronous development model

19 Asynchronous execution Ideal world Task A Task B Task C Real world Blocking I/O Asynchronous execution 19 Async programming model Non-blocking IO Task-based concurrency

20 Asynchronous, welcome to Hollywood Don t wait, we will call you... Synchronous public int compute(int a, int b) { return...; } int res = compute(1, 2); Asynchronous public void compute(int a, int b, Handler<AsyncResult<Integer>> h){ //... handler.handle(i); } 20 compute(1, 2, res -> { // Called with the // async result });

21 Callbacks for notifications and async actions Web server example vertx.createhttpserver().requesthandler(req -> // Async reaction req.response().end("reactive greetings") ).listen(8080, ar -> { // Async operation //... }); 21

22 Callbacks lead to... Reality check client.getconnection(conn -> { if (conn.failed()) {/* failure handling */} else { SQLConnection connection = conn.result(); connection.query("select * from PRODUCTS", rs -> { if (rs.failed()) {/* failure handling */} else { List<JsonArray> lines = rs.result().getresults(); for (JsonArray l : lines) { System.out.println(new Product(l)); } connection.close( done -> { if (done.failed()) {/* failure handling */} }); } }); } }); 22

23 Reactive Programming... Tame the asynchronous

24 Are you an Excel user? My Expense Report 24 Lunch 15 Coffee 25 Drinks 45 Total 85

25 Are you an Excel user? My Expense Report 25 Lunch 15 Coffee 25 Drinks 45 Total =sum(b2:b4) Observe

26 Streams (a.k.a Observables, Flowables, Publishers) My Expense Report My Expense Report Lunch 15 Coffee 0 Drinks 0 Total Lunch 15 My Expense Report Coffee 25 Lunch 15 Drinks 45 Coffee 25 Total 85 Drinks 0 Total 40 tim e

27 Reactive Programming Publisher and Subscriber

28 Reactive Extension - RX Java 2 Flowable<Integer> obs1 = Flowable.range(1, 10); Flowable<Integer> obs2 = obs1.map(i -> i + 1); Flowable<Integer> obs3 = obs2.window(2).flatmap(mathflowable::sumint); obs3.subscribe( i -> System.out.println("Computed " + i) ); 28

29 Reactive types & Asynchronous Flowable / Observable - Stream of data, Async reaction d d d X? Bounded or unbounded stream of values Data, Error, End of Stream Singles / Maybe - Async operation Stream of one value Data, Error, Completion (maybe) Single Maybe d e d X Completables - Async operation (no return) Stream without a value Completion, Error 29 X e

30 Reactive types & Asynchronous C client.rxgetconnection() // Single(async op).flatmappublisher(conn -> conn.rxquerystream( "SELECT * from PRODUCTS").flatMapPublisher(SQLRowStream::toFlowable).doAfterTerminate(conn::close) ) // Flowable of Rows.map(Product:: new) // Flowable of Products.subscribe(System. out::println); 30 flatmap flatmap Q R R P P map

31 Back pressure & Reactive streams

32 What if the processing can t keep up Streams.getStream().observeOn(Schedulers.computation()).map(i -> { Thread.sleep(100); // I'm slow... return i.getlong("id"); }).blockingsubscribe( x -> System.out.println("Processed:" + x), Throwable::printStackTrace, () -> System.out.println("Completed!") ); 32 J Stream J map Slow processing I J I J

33 Buffer to handle small bumps Streams.getStream().onBackPressureBuffer(10000).observeOn(Schedulers.computation()).map(i -> { Thread.sleep(100); // I'm slow... return i.getlong("id"); }).blockingsubscribe( x -> System.out.println("Processed:" + x), Throwable::printStackTrace, () -> System.out.println("Completed!") ); 33 Stream J J J buffer Slow processing I I J

34 Drop when data loss is acceptable Streams.getStream().onBackPressureDrop().observeOn(Schedulers.computation()).map(i -> { Thread.sleep(100); // I'm slow... return i.getlong("id"); }).blockingsubscribe( x -> System.out.println("Processed:" + x), Throwable::printStackTrace, () -> System.out.println("Completed!") ); 34 Stream Slow processing /dev/null J J I J I J

35 Control Flow and Reactive Streams Streams.getStreamWithBackPressure().observeOn(Schedulers.computation()).map(i -> { Thread.sleep(100); // I'm slow... return i.getlong("id"); }).blockingsubscribe( x -> System.out.println("Processed:" + x), Throwable::printStackTrace, () -> System.out.println("Completed!") ); Publisher J J request(x) Subscriber I Reactive Stream API - Java 9 Flow J I J

36 Back to our database access client.rxgetconnection() // Single(async op).flatmappublisher(conn -> conn.rxquerystream( "SELECT * from PRODUCTS").flatMapPublisher(SQLRowStream:: toflowable).doafterterminate(conn::close) ) // Flowable of Rows.map(Product:: new) // Flowable of Products.subscribe(p -> { // do something slow, but it s fine }); 36 request(x)

37 Eclipse Vert.x Unleash your reactive superpowers VERT.X

38 Asynchronous non-blocking Eclipse Vert.x development models Designed with reactive in mind Vert.x is a toolkit to build distributed and reactive systems Microservice, Web applications, IOT, API Gateway, High-volume event processing, Full-blown backend message bus. 38 Simplified concurrency (event loop) (Clustered) event bus Polyglot (java, scala, kotlin, groovy, ruby, ceylon )

39 Vert.x - the all-in-one toolkit Reactive A software showing responses to stimuli Reactive Systems 39 Reactive Streams Reactive Programming

40 Taming the asynchronous Distributed systems done right My Application Some Software 40 Some Services

41 Reactive Web Application private void add(routingcontext rc) { String name = rc.getbodyasstring(); database.insert(name) // Single (async).subscribe( () -> rc.response().setstatuscode(201).end(), rc::fail ); } private void list(routingcontext rc) { HttpServerResponse response = rc.response().setchunked(true); database.retrieve() // Flowable<Product> (async).subscribe( p -> response.write(json.encode(p) +" \n\n"), rc::fail, response::end); } 41 My Application

42 Orchestrating remote interactions Sequential composition WebClient pricer =... HttpServerResponse response = rc.response().setchunked(true); database.retrieve() // For each row call....flatmapsingle(p -> webclient.get("/prices/" + p.getname()).rxsend().map(httpresponse::bodyasjsonobject).map(json -> p.setprice(json.getdouble("price"))) ).subscribe( p -> response.write(json.encode(p) + " \n\n"), rc::fail, response::end); 42 My Application URL

43 Orchestrating remote interactions Event Bus WebClient pricer =... HttpServerResponse response = rc.response().setchunked(true); database.retrieve() // For each row call....flatmapsingle(p -> vertx.eventbus().<jsonobject>rxsend("pricer", p.getname()).map(message::body).map(json -> p.setprice(json.getdouble("price"))) ).subscribe( p -> response.write(json.encode(p) + " \n\n"), rc::fail, response::end); 43 My Application Address

44 Push data using event bus bridges Web Socket, SSE... String name = rc.getbodyasstring().trim(); database.insert(name).flatmap(...).subscribe( p -> { String json = Json.encode(p); rc.response().setstatuscode(201).end(json); vertx.eventbus().publish("products", json); }, rc::fail); 44 SockJS My Application

45 Executing several operations concurrently database.insert(name).flatmap(p -> { Single<Product> price = getpriceforproduct(p); Single<Integer> audit = sendactiontoaudit(p); return Single.zip(price, audit, (pr, a) -> pr); }).subscribe( p -> { String json = Json.encode(p); rc.response().setstatuscode(201).end(json); vertx.eventbus().publish("products", json); }, rc::fail); 45 My Application

46 Reactive => Build better systems

47 Welcome to an asynchronous world Reactive A software showing responses to stimuli 47 Reactive Systems Reactive Streams Responsive distributed systems Back-Pressure protocol Reactive Programming Asynchronous development model

48 Building Reactive Microservices in Java Use Eclipse Vert.x to build reactive microservices: Explore the elements of reactive microservices and learn how Vert.x works. Build and consume a single microservice to understand how messaging improves its reactiveness. Create an entire microservices system, using stability and resilience patterns to manage failures. Download the book from: clement@apache.org