[Jean-Baptiste Clion] Microservices in Google App Engine

Presentation from GDG DevFest Ukraine 2016.
Learn more at: https://devfest.gdg.org.ua

Transcript

1. Microservices in Google App Engine A Journey to resilience Jean-Baptiste Clion Founder @ Apeser | GDG Basel Founder | Google Developer Expert
2. x 1. App Engine Evolution 2. Microservices paradigm 3. Microservices on GAE
3. We will focus on production ready solution No alpha or beta here.
4. x 1. App Engine Evolution
5. App Engine Blog, Monday 7th April, 2008
6. App Engine Blog, April 7th, 2008 The goal is to make easy to get started with a new app, and then make it easy to scale when that app reaches the point where it receiving significant traffic and has millions of users.
7. Feature (2008) Behavior (Dynamic only) Deadlines 60-second deadline requests |10-minute for tasks Background threads Not allowed Residence Evicted from memory Startup and shutdown Instances created on demand | Turned down on idle Instance addressability Instances are anonymous Scaling Scales the number of instances automatically Free daily usage quota 28 instance-hours
8. App Engine ecosystem services • Datastore • Task queue • Cron tasks • User authentication • Messaging (XMPP + Email) • URL Fetch
9. What’s that means ?
10. Google I/O 2011 , Tuesday 10th May, 2011
11. Moscone Center San-Francisco Introducing Google App Engine Backends
12. 1. Split big problems into small pieces 2. Fault-tolerance 3. Horizontal scaling 4. NoSQL scalable store (With namespaces) 5. Web serving (Use HTTP protocol) App Engine Philosophy [BACKEND INTRODUCED] Justin Haugh, Google I/O 2011
13. Feature (2011) Behavior (Backend Residents) Deadlines Requests to backends can run indefinitely. Background threads Allowed Residence Possible resident instances, state is preserved. Startup and shutdown Evicted on usage patterns or manually stopped Instance addressability Instances are individually addressable via URL Scaling Configure number or maximum number of instances Free daily usage quota 28 instance-hours
14. GCP Blog, Wednesday 17th July, 2013
15. 1. Split application in different logical components. 2. Isolated settings, deployments & versions. 3. Adjust instances performances & optimize costs. Modules - Replacing backends - Freedom
16. Nowadays…. July 2016
17. July 2016 Modules are becoming Services The same but different
18. Overview 1. Split big problems into small pieces 2. Autoscaling & Fault Tolerant 3. Rely on simple HTTP(S) communication. 4. NoSQL scalable & isolable storage 5. Use automated deployment processes.
19. Split big problems, fault-tolerance, data isolation, communicating via HTTP…. Hey! Sounds like... Never mind...
20. x 2. Microservices paradigm
21. Microservices Breaking down an application into smaller composable pieces which work together. Each component is developed separately, and the application is then the sum of its constituent components.
22. Why Microservices?
23. Netflix offices, Somewhere in 2008
24. 09:15 AM A quiet and peaceful morning at Netflix
25. 09:30 AM Suddenly, out of nowhere, a semicolon is missing….
26. 09:45 AM Netflix website down
27. How? Why? Who is in charge?
28. The monolith
29. Divide & Conquer
30. Stateless instances Data isolation Scaling on Y axis API Gateway Consistent Code Maturity Async HTTP communication Divide & Conquer with microservices
31. x 3. Microservices on GAE
32. Code maturity level consistency Your services will follow the same set of coding practices. Updating your practices implies to update your source code. This will keep cohesion across your solution.
33. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? 5 8 Carts data Orders data proceed order 1 12 2 9 3 4 10 11 6 7 0 13 Different code maturity level V1. Product ID -> Long
34. Different code maturity level Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? 5 8 Carts data Orders data proceed order 1 12 2 9 3 4 10 11 6 7 0 13 V1. Product ID -> Long V2. Product ID -> String
35. Different code maturity level V1. Product ID -> Long V2. Product ID -> String Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? 5 8 Carts data Orders data proceed order 1 12 2 9 3 4 10 11 6 7 0 13
36. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? 5 8 Carts data Orders data proceed order 1 12 2 9 3 4 10 11 6 7 0 13 Same code maturity level V2. Product ID -> String V2. Product ID -> String
37. App Engine Feature Services versions
38. Scaling on Y axis You code will have a functional decomposition by following your business logic [Services]. Your solution will then scale on services instead of entire application. This will have a drastic positive impact on resources.
39. Y Axis Functional decomposition Scale by splitting different things X Axis Horizontal duplication Scale by cloning Z Axis Data partitioning Scale by splitting similar things The Scale cube
40. Frontend Products Frontend Orders products existing? Products data Orders data proceed order Frontend Carts cart valid? Carts data X Scaling - Scale entire application
41. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Carts data Orders data proceed order Frontend Products products existing? Products data Frontend Carts cart valid? Carts data X Scaling - Scale entire application Frontend Orders Orders data proceed order
42. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Carts data Orders data proceed order Frontend Products products existing? Products data Frontend Carts cart valid? Carts data Frontend Orders Orders data proceed order X Scaling - Scale entire application
43. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Products data Orders data proceed order Y Scaling - Scale service only
44. products existing? Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Products data Orders data proceed order Y Scaling - Scale service only
45. products existing? products existing? Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Products data Orders data proceed order Y Scaling - Scale service only
46. App Engine Feature Native autoscaling
47. Async HTTP communication Synchronous communications via REST or asynchron via message broker. Simple HTTP communications allow flexibility and interoperability between services.
48. App Engine Feature Synchronous - URLFetch Asynchronous - Task Queue
49. Stateless instances Do not persist information between requests. This will allow your instances to scale and increase the amount of worked produced without data inconsistencies.
50. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? 5 8 Carts data Orders data proceed order 1 12 2 9 3 4 10 11 6 7 0 13 Stateful servers
51. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? 5 8 Carts data Orders data proceed order 1 12 2 9 3 4 10 11 6 7 0 13 Product $30 Caching Stateful servers
52. products existing? Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? 5 8 Carts data Orders data proceed order 1 12 2 9 3 4 10 11 6 7 0 13 Product $10 Product $30 Caching Stateful servers
53. products existing? products existing? Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? 5 8 Carts data Orders data proceed order 1 12 3 4 10 11 6 7 2 9 0 13 Product $20 Product $10 Product $30 Caching Stateful servers
54. products existing? products existing? Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? 5 8 Carts data Orders data proceed order 1 12 2 9 3 4 10 11 6 7 0 13 Product $20 Product $10 Product $30 Caching Stateful servers
55. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? 5 8 Carts data Orders data proceed order 1 12 2 9 3 4 10 11 6 7 0 13 Stateless servers Product $30
56. products existing? Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? 5 8 Carts data Orders data proceed order 1 12 2 9 3 4 10 11 6 7 0 13 Product $30 Stateless servers
57. products existing? products existing? Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? 5 8 Carts data Orders data proceed order 1 12 2 9 3 4 10 11 6 7 0 13 Product $30 Stateless servers
58. App Engine Feature Datastore & Memcache are decoupled of GAE instances
59. Data isolation Each service has an isolated and independent persisted data set. Isolation globally reduce pressure on a single database and avoid overall collapsing due to data interdependencies.
60. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Carts data Orders data proceed order Non isolated datasets
61. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Carts data Orders data proceed order Non isolated datasets
62. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Carts data Orders data proceed order Non isolated datasets
63. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Carts data Orders data proceed order Have access to unnecessary data Non isolated datasets
64. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Carts data Orders data proceed order Have access to unnecessary data Code will slowly be implemented in other service. Non isolated datasets
65. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Carts data Orders data proceed order Have access to unnecessary data Code will slowly be implemented in other service. Non isolated datasets
66. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Carts data Orders data proceed order Non isolated datasets
67. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Carts data Orders data proceed order Non isolated datasets
68. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Carts data Orders data proceed order Non isolated datasets
69. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Carts data Orders data proceed order Non isolated datasets
70. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Carts data Orders data proceed order Non isolated datasets
71. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? 5 8 Carts data Orders data proceed order 1 12 2 9 3 4 10 11 6 7 0 13 Isolated data
72. API Gateway System single entry point, the gateway will control and execute the different services required by your client. It will simplify communications and avoid arbitrary API calls.
73. Frontend Products Frontend Carts Frontend Orders products existing? Products data cart valid? Carts data Orders data proceed order No Api Gateway
74. Frontend Products Frontend Carts Frontend Orders Check product Products data Check cart Carts data Orders data proceed order No Api Gateway 1. Call - Check product 2. Call - Check cart 3. Call - proceed order
75. Frontend Products Frontend Carts Frontend Orders Check product Products data Check cart Carts data Orders data proceed order Api Gateway 1. Call - Check product 2. Call - Check cart 3. Call - proceed order Check out Proxy
76. App Engine Feature Modular Services
77. Overview 1. Split large problem in small dedicated modules. 2. Modules are low coupled to ease maintenance. 3. Rely on simple HTTP(S) communication. 4. Data isolation as best practice. 5. Use automated deployment processes.
78. Wait a minute! Split big problems, fault-tolerance, data isolation, communicating via HTTP….
79. Microservices architecture in 2008 !!!
80. Lightweight Event-based Microservices Google Cloud Functions is a lightweight, event-based, asynchronous compute solution that allows you to create small, single-purpose functions that respond to cloud events without the need to manage a server or a runtime environment. CLOUD FUNCTIONS ALPHA
81. @JBaptisteClion News, Questions, Time paradoxes? Contact me on
84. Questions & Answers
85. Jean-Baptiste Clion Founder @Apeser | GDG Basel Founder | GDE GCP jbclion +JeanBaptisteClionBasel @JBaptisteClion