Serverless for Developers: Tips for Your Next App

ServerlessConf, London, October 28th, 2016

Serverless architectures simplify operations and maintenance of production applications, abstracting from the underlining infrastructure and OS. But what about development? We are now two years in with tools such as AWS Lambda and what are the best practices to share from a development perspective? In this session you will see some of the key points that can empower a serverless developer: from event-driven design, to simplification (less is more, especially for a small team), extending the functionalities of the platform (per prefix stats for S3 buckets, geohash for DynamoDB tables, API authentication, and so on), integrating new products.

Transcript

1. Serverless for Developers: Tips for Your Next App Danilo Poccia, Technical Evangelist @danilop danilop
3. Credit: Gerry Cranham/Fox Photos/Getty Images http://www.telegraph.co.uk/travel/destinations/europe/united-kingdom/england/london/galleries/The-history-of-the-Tube-in-pictures-150-years-of-London-Underground/1939-ticket-examin/
4. Credit: Gerry Cranham/Fox Photos/Getty Images http://www.telegraph.co.uk/travel/destinations/europe/united-kingdom/england/london/galleries/The-history-of-the-Tube-in-pictures-150-years-of-London-Underground/1939-ticket-examin/ 1939 London Underground
5. AWS IoT Kinesis Stream AWS Lambda DynamoDB Table S3 Bucket
6. Dev + Ops
7. Dev + Ops + Event Driven
8. Name Email Create New User Submit Check the “Name” syntax is right (only letters and spaces) Check the “Email” syntax is right (something@some.domain) Create the new user using provided Name and Email User Interface Actions
9. Name Email Create New User Submit Check the “Name” syntax is right (only letters and spaces) Check the “Email” syntax is right (something@some.domain) Create the new user using provided Name and Email User Interface Actions Observers Target Action
10. Generate thumbnails Keep another database in sync Process streaming data Services Functions Platform Target Action Kinesis Stream DynamoDB Table S3 Bucket
11. area = length x width
12. area = length x width Procedural Programming “This is a function!”
13. area = length x width Procedural Programming “This is a function!” Reactive Programming “This is data binding!”
14. A B Event-Driven Design For Services A triggers B or better B is caused by A
15. A B C D New Services C and D are triggered caused by B
16. A B C D E F ? ? What is causing E and F?
17. A B C D E F New Services (E, F) are triggered by C
18. A B C D E F It can be cyclic Think of acknowledgements
19. A B C D E F Each service has local visibility
20. A B C D E F What I need to know (input events) 1
21. A B C D E F What I need to do (internal logic) 2
22. A B C D E F Who I need to notify (output events) 3
23. A B C D E F Who I need to notify (output events) What I need to know (input events) What I need to do (internal logic) 1 2 3
24. Distributed Data Flow Vs Centralised Workflow
25. Choreography Vs Orchestration
26. Distributed Systems
27. Monolith Service Service μ μ μ space (distribution, mobility) time (concurrency, latency) μ μ μ μ “Distributed in Time and Space” [Jonas Bonér]
28. “…a diagram of two microservices and their shared database” Data should drive the boundaries
29. A C I D tomic onsistent isolated urable
30. A C I D tomic onsistent isolated urable A C I D ssociative ommutative dempotent istributed ACID 2.0
31. A B C D E F Event-Driven Interactions replacing Distributed Transactions using Eventual Consistency
32. A B C D E F Event-Driven Interactions replacing Distributed Transactions using Eventual Consistency Differentiate reads (queries) from writes (commands), as in the Command Query Responsibility Segregation (CQRS) pattern command query
33. Customers Orders Warehouse Deliveries Payments Catalog Event-Driven Interactions replacing Distributed Transactions using Eventual Consistency
34. Scale from Prototype to Production
35. Responsive Resilient Elastic Message Driven The Reactive Manifesto
36. The Reactive Manifesto Asynchronous communication is not enforced by serverless architectures Responsive Resilient Elastic Message Driven
37. Serverless Architectural Patterns
38. AWS Lambda User Client Application Back End for a Web or Mobile App Calling the Lambda API directly Web or Mobile Back End DynamoDB Table S3 Bucket Amazon Cognito AWS API
39. AWS Lambda Amazon Cognito User Client Application OAuth JWT Amazon API Gateway Back End for a Web or Mobile App Using an API Gateway for authentication and decoupling Web or Mobile Back End Via API Gateway DynamoDB Table S3 Bucket HTTPS
40. Using WebSockets instead of Web API to talk to the Back End WebSockets Back End AWS Lambda User Client Application AWS IoT DynamoDB Table S3 Bucket Amazon Cognito MQTT or Device Shadow MQTT over
 WebSockets
41. Server ! Client { "id": "...", "html": "..." } { "js": "..." } Client ! Server { "cmd": "...", "data": "..." } MQTT Topics myapp/client/in/{ID} myapp/client/out/{ID} Cognito Identity ID ! {ID} 1 2 3 4 AWS Lambda User Client Application AWS IoT DynamoDB Table S3 Bucket Amazon Cognito MQTT or Device Shadow MQTT over
 WebSockets
42. Possible optimisations depending on your data flow and your acceptable latency Data Source Kinesis Stream Kinesis Firehose Kinesis Analytics (augment & aggregate) AWS Lambda (batch) S3 Bucket AWS Lambda (file) Data Streams Processing Reference Data
43. AWS Lambda (batch) Amazon Cognito User Client Application Buffering Requests to the Back End to reduce Lambda invocations Kinesis Stream Buffered Back End For Mobile Amazon SNS DynamoDB Table S3 Bucket
44. AWS Lambda (batch) Amazon Cognito User Client Application Buffering Requests to the Back End to reduce Lambda invocations Kinesis Stream Buffered & Aggregated
 Back End For Mobile Amazon SNS DynamoDB Table S3 Bucket Kinesis Analytics (aggregate) Kinesis Stream
45. AWS Lambda (batch) User Client Application Using WebSockets instead of Web API to talk to the Back End Buffering Requests to reduce Lambda invocations AWS IoT Kinesis Stream Buffered Back End For Everything MQTT or Device Shadow Amazon Cognito DynamoDB Table S3 Bucket
46. User Client Application Using WebSockets instead of Web API to talk to the Back End Buffering Requests to reduce Lambda invocations Storing all customer interactions for further analysis AWS IoT Kinesis Firehose Buffered Back End Saving Interactions MQTT or Device Shadow Amazon Cognito AWS Lambda (file) S3 Bucket DynamoDB Table
47. AWS Lambda DynamoDB Table User Client Application Amazon API Gateway (HTTPS) S3 Bucket (Website Hosting Routing Rules) For example, to generate image thumbnails on the first request, S3 LifeCycle Rules can delete content after N days Dynamic Cache Amazon Cognito HTTP Redirect Dynamically Build Content
48. To receive events from everywhere on the Internet Webhooks are callbacks for the Web AWS Lambda Amazon API Gateway Internet Service HTTP(S) GET or POST to https://domain/SOURCE/RANDOM-HOOK Webhook Other AWS Services Other public APIs AWS KMS Secrets & Credentials
49. To generate events for repositories that don’t support AWS Lambda natively Log Monitor AWS Lambda AWS Lambda AWS Lambda Amazon SNS Kinesis Stream Log Files External Application Log Monitor CloudWatch Logs or Metrics AWS Lambda
50. S3 object auto compress S3 bucket per prefix statistics DynamoDB provisioned throughput dynamic scaling DynamoDB item support for geohash Relational / NoSQL data synchronisation Service discovery . . . Extending Services with Your Own Functionalities
51. Serverless === Architecture Simplification
52. Serverless === Architecture Simplification For example, let’s build a Media Sharing App
53. Client Application Get Content Index Get Content and Thumbnails Upload Content with Metadata Update Content Metadata Get Content Metadata Files (Multimedia) Database Build Thumbnails Update Content Index Extract and Update Metadata Functions User Resources Basic features M edia Sharing App
54. Client Application Get Content Index Get Content and Thumbnails Upload Content with Metadata Update Content Metadata Get Content Metadata Files (Multimedia) Database Build Thumbnails Update Content Index Extract and Update Metadata Event: new or updated file Event: new or updated content metadata Functions User Resources Event-driven design M edia Sharing App
55. Client Application Get Content Index Get Content and Thumbnails Upload Content with Metadata Update Content Metadata Get Content Metadata Files (Multimedia) Database Build Thumbnails Update Content Index Extract and Update Metadata Event: new or updated file Event: new or updated content metadata Functions User Resources Understanding the overall flow M edia Sharing App
56. Client Application Get Content Index (S3 API) Get Content and Thumbnails (S3 API) Upload Content
 with Metadata (S3 API) Update Content Metadata (DynamoDB API) Get Content Metadata (DynamoDB API) S3 Bucket DynamoDB Table Build Thumbnails (Lambda function) Update Content Index (Lambda function) Extract and Update Metadata (Lambda function) Event: new or updated file Event: new or updated content metadata User Choosing the technology M edia Sharing App
57. Client Application Get Content Index (S3 API) Get Content and Thumbnails (S3 API) Upload Content
 with Metadata (S3 API) Update Content Metadata (DynamoDB API) Get Content Metadata
 (DynamoDB API) S3 Bucket DynamoDB Table Build Thumbnails (Lambda function) Update Content Index (Lambda function) Extract and Update Metadata (Lambda function) Event: new or updated file Event: new or updated content metadata User Amazon Cognito Defining the security model M edia Sharing App
58. Client Application Get Content Index (S3 API) Get Content and Thumbnails (S3 API) Upload Content
 with Metadata (S3 API) Update Content Metadata (DynamoDB API) Get Content Metadata
 (DynamoDB API) S3 Bucket DynamoDB Table Build Thumbnails (Lambda function) Update Content Index (Lambda function) Extract and Update Metadata (Lambda function) Event: new or updated file Event: new or updated content metadata User Amazon Cognito From the features viewpoint… M edia Sharing App
59. Client Application GET Object S3 API PUT Object S3 API UpdateItem DynamoDB API GetItem
 DynamoDB API S3 Bucket DynamoDB Table buildThumbnails (Lambda function) updateContentIndex (Lambda function) extractAndUpdateMetadata (Lambda function) Event: new or updated file Event: new or updated content metadata User Amazon Cognito Amazon S3 Amazon DynamoDB AWS Lambda …To the technology viewpoint M edia Sharing App
60. Client Application GET Object S3 API PUT Object S3 API UpdateItem DynamoDB API GetItem
 DynamoDB API S3 Bucket DynamoDB Table buildThumbnails (Lambda function) updateContentIndex (Lambda function) extractAndUpdateMetadata (Lambda function) Event: new or updated file Event: new or updated content metadata User Amazon Cognito Amazon S3 Amazon DynamoDB AWS Lambda Some events are correlated M edia Sharing App
61. Client Application GET Object S3 API PUT Object S3 API UpdateItem DynamoDB API GetItem
 DynamoDB API S3 Bucket DynamoDB Table contentUpdated (Lambda function) buildThumbnails extractAndUpdateMetadata updateContentIndex (Lambda function) Event: new or updated file User Amazon Cognito Amazon S3 AWS Lambda Amazon DynamoDB Event: new or updated content metadata Some functions can be tied together M edia Sharing App
62. Client Application GET Object S3 API PUT Object S3 API UpdateItem DynamoDB API GetItem
 DynamoDB API S3 Bucket DynamoDB Table contentUpdated (Lambda function) buildThumbnails extractAndUpdateMetadata updateContentIndex (Lambda function) Event: new or updated file User Amazon Cognito Amazon S3 AWS Lambda Amazon DynamoDB Event: new or updated content metadata What if I want to delete content? M edia Sharing App
63. Client Application GET Object S3 API PUT Object S3 API UpdateItem DynamoDB API GetItem
 DynamoDB API S3 Bucket DynamoDB Table contentUpdated (Lambda function) buildThumbnails extractAndUpdateMetadata deleteMetadata updateContentIndex (Lambda function) Event: new, updated, or deleted file User Amazon Cognito Amazon S3 AWS Lambda Amazon DynamoDB DELETE Object S3 API Event: new or updated content metadata REST to the rescue, the events flow is almost unchanged M edia Sharing App
64. Client Application GET Object S3 API PUT Object S3 API UpdateItem DynamoDB API GetItem
 DynamoDB API S3 Bucket DynamoDB Table contentUpdated (Lambda function) buildThumbnails extractAndUpdateMetadata deleteMetadata updateContentIndex (Lambda function) Event: new, updated, or deleted file User Amazon Cognito Amazon S3 AWS Lambda Amazon DynamoDB DELETE Object S3 API Event: new or updated content metadata From 8 functions to 2, using 5 native API calls M edia Sharing App
65. Serverless Architectures are not just simplifying Operations
66. Developers can benefit from: event-driven design architectural patterns integrating new products architecture simplification customization of the platform Serverless Architectures are not just simplifying Operations
67. Build Apps With Services, Not Servers
68. Thank you @danilop danilop