Decomposing applications for deployability and scalability (SpringOne China 2012)


Published on

A look at how decomposing applications improves deployability and scalability

Published in: Technology
  • Be the first to comment

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Decomposing applications for deployability and scalability (SpringOne China 2012)

  1. 1. Decomposing applications for deployability and scalabilityChris Richardson,Author of POJOs in Action, Founder of the original @crichardson
  2. 2. Presentation goal How decomposing applications improves deployability and scalability and How Cloud Foundry helps 2
  3. 3. About Chris 3
  4. 4. (About Chris) 4
  5. 5. About Chris() 5
  6. 6. About Chris 6
  7. 7. About Chris 7
  8. 8. vmc push About-Chris Developer Advocate Signup at 8
  9. 9. Agenda§The (sometimes evil) monolith§Decomposing applications into services§How do services communicate?§Presentation layer design§How Cloud Foundry helps 9
  10. 10. Let’s imagine you are building an e-commerce application 10
  11. 11. Traditional web application architecture WAR StoreFrontUI Accounting Service MySQL Browser Apache InventoryService Database Shipping Service Simple to Tomcat develop test deploy scale 11
  12. 12. But there are problems with a monolithic architecture 12
  13. 13. Users expect a rich, dynamicand interactive experience h oug en HTTP Request od ’t go Browser isn Java Web Application re ctu HTML/Javascript e rc hit I a ty le U s Old Real-time web ≅ NodeJS 13
  14. 14. Intimidates developers 14
  15. 15. Obstacle to frequent deployments§ Need to redeploy everything to change one component§ Interrupts long running background (e.g. Quartz) jobs§ Increases risk of failure Fear of change§ Updates will happen less often§ e.g. Makes A/B testing UI really difficult 15
  16. 16. Overloads your IDE and container Slows down development 16
  17. 17. Obstacle to scaling development Accounting team E-commerce Engineering application Shipping team 17
  18. 18. Obstacle to scaling development WAR UI team StoreFrontUI Accounting team Accounting Inventory team InventoryService Shipping team Shipping 18
  19. 19. Obstacle to scaling development But the I want to update backend is not the UI working yet! Lots of coordination and communication required 19
  20. 20. Requires long-term commitment to a technology stack 20
  21. 21. Agenda§The (sometimes evil) monolith§Decomposing applications into services§How do services communicate?§Presentation layer design§How Cloud Foundry helps 21
  22. 22. 22
  23. 23. The scale cube Y axis - functional decomposition ila ng Scale by i m ni g s itio r splitting tin rt lit pa different things sp a gs y dat th ale is - Sc ax in b X axis Z - horizontal duplication 23
  24. 24. Y-axis scaling - application level WAR StoreFrontUI Accounting Service InventoryService Shipping Service 24
  25. 25. Y-axis scaling - application level accounting web application Accounting Service Store front web application inventory web application StoreFrontUI InventoryService shipping web application Shipping Service Apply X axis cloning and/or Z axis partitioning to each service 25
  26. 26. Partitioning strategies§Partition by verb, e.g. shipping service§Partition by noun, e.g. inventory service§Single Responsibility Principle§Unix utilities - do one focussed thing well Something of an art 26
  27. 27. Real world examples Between 100-150 services are accessed to build a page. eBaySDForum2006-11-29.pdf 27
  28. 28. There are drawbacks 28
  29. 29. ComplexitySee Steve Yegge’s Google Platforms Rant re 29
  30. 30. Multiple databases =Transaction management challenges 30
  31. 31. When to use it? In the beginning: •You don’t need it •It will slow you down Later on: •You need it •Refactoring is painful 31
  32. 32. But there are many benefits§Scales development: develop, deploy and scale each service independently§Update UI independently§Improves fault isolation§Eliminates long-term commitment to a single technology stack Modular, polyglot, multi-framework applications 32
  33. 33. Two levels of architecture System-levelServicesInter-service glue: interfaces and communication mechanismsSlow changing Service-levelInternal architecture of each serviceEach service could use a different technology stackPick the best tool for the jobRapidly evolving 33
  34. 34. If services are small...§Regularly rewrite using a better technology stack§Adapt system to changing requirements and better technology without a total rewrite§Pick the best developers rather than best <pick a language> developers polyglot culture 34
  35. 35. The human body as a system 35
  36. 36. 50 to 70 billion of your cells die each day 36
  37. 37. Yet you (the system) remain you 37
  38. 38. Can we build software systems with these characteristics? DesignBeyondHumanAbilitiesSimp.pdf 38
  39. 39. Agenda§The (sometimes evil) monolith§Decomposing applications into services§How do services communicate?§Presentation layer design§How Cloud Foundry helps 39
  40. 40. Inter-service communication options§Synchronous HTTP asynchronous AMQP§Formats: JSON, XML, Protocol Buffers, Thrift, ...§Even via the database Asynchronous is preferred JSON is fashionable but binary format is more efficient 40
  41. 41. Asynchronous message-based communication wgrus-billing.war Accounting Servicewgrus-store.war wgrus-inventory.war RabbitMQ StoreFrontUI (Message InventoryService MySQL Broker) wgrus-shipping.war ShippingService 41
  42. 42. Benefits§Decouples caller from server§Caller unaware of server’s coordinates (URL)§Message broker buffers message when server is down/slow§Supports a variety of communication patterns, e.g. point-to-point, pub-sub, ... 42
  43. 43. Drawbacks§Additional complexity of message broker§Request/reply-style communication is more complex 43
  44. 44. Writing code that calls services 44
  45. 45. The need for parallelism Service B b = serviceB() Call in parallel c = serviceC() Service A Service C d = serviceD(b, c) Service D 45
  46. 46. Java Futures are a great concurrency abstraction 46
  47. 47. Using Java Futurespublic class Client { private ExecutorService executorService; private RemoteServiceProxy remoteServiceProxy; public void doSomething() throws ... { Eventually contains result Future<Integer> result = executorService.submit(new Callable<Integer>() { @Override public Integer call() throws Exception { return remoteServiceProxy.invokeRemoteService(); } }); /// Do other things When needed wait for result int r = result.get(500, TimeUnit.MILLISECONDS); System.out.println(r); 47 }}
  48. 48. Akka’s composable futures are even better 48
  49. 49. Composable Futuresval f1 = Future { ... ; 1 }val f2 = Future { ... ; 2 } Transforms Futureval f4 = * 2)assertEquals(4, Await.result(f4, 1 second)) Combines two futuresval fzip = f1 zip f2assertEquals((1, 2), Await.result(fzip, 1 second)) 49
  50. 50. Using Akka futures def callB() : Future[...] = ... def callC() : Future[...] = ... def callD() : Future[...] = ... Two calls execute in parallel val future = for { (b, c) <- callB() zip callC(); d <- callD(b, c) } yield d And then invokes D val result = Await.result(future, 1 second) Get the result of D 50
  51. 51. Spring Integration§Provides the building blocks for a pipes and filters architecture§Enables development of application components that are •loosely coupled •insulated from messaging infrastructure§Messaging defined declaratively 51
  52. 52. Handling failure Service A Service B Errors happen in distributed systems 52
  53. 53. About Netflix > 1B API calls/day 1 API call average 6 service calls Fault tolerance is essential 53
  54. 54. How to run out of threadsHTTP Request X Thread 1 X Thread 2 ... X X Thread N Execute thread Service A XService B pool Eventually all threads If service B is down then thread will be Tomcat will be blocked blocked 54
  55. 55. Use timeouts and retries Never wait forever Errors can be transient retry 55
  56. 56. Use per-dependency bounded thread poolService A Runnable 1 Task 1 Runnable 2 Task 2 Service B Runnable ... Task ... bounded queue bounded thread pool Fails fast if Limits number of service is slow or down outstanding requests 56
  57. 57. Use a circuit breaker High error rate stop calling temporarily Down wait for it to come back up Slow gives it a chance to recover 57
  58. 58. On failure Return cached data Avoid Failing Return default data Fail fast 58
  59. 59. Agenda§The (sometimes evil) monolith§Decomposing applications into services§How do services communicate?§Presentation layer design§How Cloud Foundry helps 59
  60. 60. Modular applicationChoice of presentation layer technology 60
  61. 61. NodeJS is the fashionable technology 61
  62. 62. Why NodeJS?§Familiar Javascript§High-performance, scalable event-driven, non-blocking I/O model§Compact runtime§Over 17,000 modules developed by the community§Many JavaScript client frameworks have a NodeJS counterpart, e.g. and SockJS 62
  63. 63. Why not NodeJS? a.k.a. callback hell 63
  64. 64. A modern web application Service 1 RESTful WS Node JSBrowser HTML 5 Server Application Application Service 2 Events client server ... 64
  65. 65. NodeJS - using RESTful WS and AMQP REST ServiceRESTRequests Node JS Events AMQP AMQP RabbitMQ Service 65
  66. 66. server-side var express = require(express) , http = require(http) , amqp = require(‘amqp’) ....; server.listen(8081); Handle ... connection var amqpCon = amqp.createConnection(...); io.sockets.on(connection, function (socket) { Republish as function amqpMessageHandler(message, headers, deliveryInfo) { event var m = JSON.parse(; socket.emit(‘tick’, m); }; amqpCon.queue(“”, {}, function(queue) { Subscribe to AMQP queue.bind(“myExchange”, “”); queue.subscribe(amqpMessageHandler); queue }); }); 66
  67. 67. - client side<html><body>The event is <span data-bind="text: ticker"></span><script src="/"></script><script src="/knockout-2.0.0.js"></script><script src="/clock.js"></script></body> Bind to model Connect to</html> server clock.jsvar socket = io.connect(location.hostname);function ClockModel() { self.ticker = ko.observable(1); Subscribe to tick socket.on(tick, function (data) { event self.ticker(data); });}; Update modelko.applyBindings(new ClockModel()); 67
  68. 68. Agenda§The (sometimes evil) monolith§Decomposing applications into services§How do services communicate?§Presentation layer design§How Cloud Foundry helps 68
  69. 69. Original architecture WAR StoreFrontUI Accounting Service MySQLBrowser Apache InventoryService Database Shipping Service Tomcat 69
  70. 70. Modern architecture p, lo Desktop Browser Native Mobile application HTML5 mobile application ve StoreUI StoreUI StoreUI st de ? is NodeJSAsynchronous, scalable NodeJS Javascript communication StoreUI th te e w RabbitMQ oy do Spring/Scala web pl application Inventory Service Shipping Service Standalone Inventory Service Shipping Service “headless” Spring/ ow Java applications de H MySQL Customer Redis Inventory Mongo Order d Billing Service Database Database Database an 70
  71. 71. Traditional tools: monolithic applications 71
  72. 72. Developing modular apps is more difficult§Many more moving parts to manage •Platform services: SQL, NoSQL, RabbitMQ •Application services: your code§Who is going to setup the environments: •the developer sandbox? •... •QA environments? But Cloud Foundry helps... 72
  73. 73. Easy polyglot application deployment and serviceprovisioning OSS community vFabric Postgres Ap p lica Private   o Data Services Clouds   n  S erv ice  In ter fac vFabric e RabbitMQTM Public Clouds Msg Services Micro Clouds Other Services Additional partners services …
  74. 74. Creating a platform service instance$ vmc create-service mysql --name mysql1Creating Service: OK$ vmc services......=========== Provisioned Services ============+-------------+---------+| Name | Service |+-------------+---------+| mysql1 | mysql |+-------------+---------+
  75. 75. Multi-application manifest - part 1---applications: inventory/target: Path to application name: inventory url: framework: name: spring info: mem: 512M description: Java SpringSource Spring Application exec: mem: 512M instances: 1 services: Required platform services si-rabbit: type: :rabbitmq si-mongo: type: :mongodb si-redis: type: :redis 75
  76. 76. Multi-application manifest - part 2store/target: Path to application name: store url: framework: name: spring info: mem: 512M description: Java SpringSource Spring Application exec: mem: 512M instances: 1 services: Required platform services si-mongo: type: :mongodb si-rabbit: type: :rabbitmq 76
  77. 77. One command to create platform services and deployapplication$ vmc pushWould you like to deploy from the current directory? [Yn]:Pushing application inventory...Creating Application: OKCreating Service [si-rabbit]: OKBinding Service [si-rabbit]: OKCreating Service [si-mongo]: OKBinding Service [si-mongo]: OKCreating Service [si-redis]: OKBinding Service [si-redis]: OKUploading Application: Checking for available resources: OK vmc push: Processing resources: OK •Reads the manifest file Packing application: OK Uploading (12K): OK •Creates the required platform servicesPush Status: OK •Deploys all the applicationsStaging Application inventory: OKStarting Application inventory: OKPushing application store...Creating Application: OKBinding Service [si-mongo]: OKBinding Service [si-rabbit]: OKUploading Application: Checking for available resources: OK Processing resources: OK Packing application: OK 77
  78. 78. Micro Cloud Foundry: new developer sandbox App Instances Services Open source Platform as a Service project 10.04 A PaaS packaged as a VMware Virtual Machine Use as a developer sandbox • Use the services from Junit integration tests • Deploy your application for functional testing • Remote debugging from STS 78
  79. 79. Using Caldecott to tunnel into your services 79
  80. 80. Caldecott = TCP over HTTP native protocol native HTTP Service Caldecott protocol Caldecott gem Service client Port application NNNYour computer Cloud Foundry 80
  81. 81. Using Caldecott…$ vmc tunnel1: mysql-135e02: mysql1Which service to tunnel to?: 2Password: ********Stopping Application: OKRedeploying tunnel application caldecott.Uploading Application: Checking for available resources: OK Packing application: OK Uploading (1K): OKPush Status: OKBinding Service [mysql1]: OKStaging Application: OKStarting Application: OKGetting tunnel connection info: OKService connection info: username : uMe6Apgw00AhS password : pKcD76PcZR7GZ name : d7cb8afb52f084f3d9bdc269e7d99ab50Starting tunnel to mysql1 on port 10000.1: none2: mysqlWhich client would you like to start?: 2
  82. 82. …Using CaldecottLaunching mysql --protocol=TCP --host=localhost --port=10000 --user=uMe6Apgw00AhS -- password=pKcD76PcZR7GZ d7cb8afb52f084f3d9bdc269e7d99ab50Welcome to the MySQL monitor. Commands end with ; or g.Your MySQL connection id is 10944342Server version: 5.1.54-rel12.5 Percona Server with XtraDB (GPL), Release 12.5, Revision 188Copyright (c) 2000, 2011, Oracle and/or its affiliates. All rights reserved.Oracle is a registered trademark of Oracle Corporation and/or itsaffiliates. Other names may be trademarks of their respectiveowners.Type help; or h for help. Type c to clear the current input statement.mysql>
  83. 83. Running JUnit test with Caldecott Configure your test code to use port + connection info 83
  84. 84. Summary 84
  85. 85. Monolithic applications are simple to develop and deployBUT have significant drawbacks 85
  86. 86. Apply the scale cube §Modular, polyglot, and scalable applications Y axis - functional decomposition §Services developed, deployed and scaled independently ngi on iti rt pa a at -d is ax Z X axis - horizontal duplication 86
  87. 87. Cloud Foundry helps Ap .js p lic a on Private    Se Data Services Clouds   rv ice  In te rfa ce Public Clouds ce rfa Msg Services te  In er vid ro d  P ou Micro Cl Clouds Other Services
  88. 88. @crichardson Questions? @cloudfoundry
  89. 89. Cloud Foundry 启动营在www.cloudfoundry.com注册账号并成功上传应用程序,即可于12月8日中午后凭账号ID和应用URL到签到处换取Cloud Foundry主题卫衣一件。89
  90. 90. iPhone5 等你拿第二天大会结束前,请不要提前离 ,将填写完整的意见反馈表投到签到处的抽奖箱内,即可参与“iPhone5”抽奖活动。90
  91. 91. Birds of a Feather 专家面对面所有讲师都会在课程结束后,到紫兰厅与来宾讨论课程上的问题91