Message Driven Architecture on AWS

Slide 1

Slide 1 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Jerome Van Der Linden (@jeromevdl) Message-Driven Architecture with AWS Worldwide Software Architecture Summit 26th of January 2021

Slide 2

Slide 2 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Who am I ? Jérôme Van Der Linden Solutions Architect @ AWS (Geneva, Switzerland) Former consultant, developer, tech lead, agile & devops coach, architect @ OCTO Technology Husband and Dad of 3 @jeromevdl

Slide 3

Slide 3 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Agenda Monolith to Microservices Microservices to Reactive Microservices Message-Driven Architecture on AWS Wrap-up

Slide 4

Slide 4 text

Slide 5

Slide 5 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Monoliths Do not throw the baby out with the bath water… - Simple to develop and refactor - Simple to test (integration, end-to-end) - Single deployment process - Single application to monitor - Consistency https://martinfowler.com/bliki/MonolithFirst.html

Slide 6

Slide 6 text

© 2021, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Confidential and Trademark Monoliths … but definitely some limits! - Strong coupling: less flexibility / agility - High impact of change - Redeploy everything - Risks of blast radius - Reliability - One failure can bring down the whole application - Rigid scaling - Often stateful - Limited to the database scalability - Development lifecycle is typically slow - Continuous deployment is difficult - Barrier to adopting new technologies

Slide 7

Slide 7 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Microservices Microservices to the rescue… - Reduced coupling - Smaller impact of changes - Independent deployments - Better reliability - Independent scaling - Smaller unit of development: - Humanely understandable - More suitable for adopting new technologies

Slide 8

Slide 8 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Microservices … but it is not that magic! - Reduced coupling - Smaller impact of changes - Independent deployments - Better reliability - Independent scaling - Smaller unit of development: - Humanely understandable - More suitable for adopting new technologies Are you sure?

Slide 9

Slide 9 text

Slide 10

Slide 10 text

Slide 11

Slide 11 text

Slide 12

Slide 12 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Synchronous communica=on Order Service Invoice Service 500 internal server error 503 service unavailable ☹

Slide 13

Slide 13 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Synchronous communica=on What’s the problem ? - No isolation between services è still coupled - Risk of timeout or cascading failures è still unreliable - What if the signature of invoice API change ? è still dependent

Slide 14

Slide 14 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Reac=ve Principles Responsive Resilient Elastic Message-Driven https://www.reactivemanifesto.org/

Slide 15

Slide 15 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Message-Driven (Asynchronous) Communica=on Order Service Invoice Service 202 accepted order created

Slide 16

Slide 16 text

Slide 17

Slide 17 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Message-Driven (Asynchronous) Communica=on Isolation between services: - Less risk of timeout è more responsive - No more cascading failure è more resilient è more responsive - What if the signature of invoice API change ? è No problem !

Slide 18

Slide 18 text

Slide 19

Slide 19 text

Slide 20

Slide 20 text

© 2021, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Confidential and Trademark Amazon Simple Queue Service (SQS) Producer SQS Queue Decouple producers from consumers. Messages polled by consumers Amazon SQS Fully managed Message Queuing service Consumers AWS Lambda Container Instance / VM Standard • At-least once delivery • Best-effort ordering Messages are stored up to 14 days.

Slide 21

Slide 21 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Amazon Simple Queue Service (SQS) Producer Consumers AWS Lambda Amazon SQS Fully managed Message Queuing service SQS Queue Decouple producers from consumers Messages polled by consumers Container Instance / VM FIFO • Exactly once delivery • Keep ordering 1 2 3

Slide 22

Slide 22 text

Slide 23

Slide 23 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Amazon Simple Queue Service (SQS) Producer Consumers AWS Lambda Amazon SQS Fully managed Message Queuing service SQS Queue Decouple producers from consumers Messages polled by consumers Container Instance / VM Dead-letter Queue If a message can’t be processed successfully it can be held in a queue

Slide 24

Slide 24 text

© 2021, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Confidential and Trademark Common paIern with SQS: Storage-first Amazon API Gateway Amazon SQS Queue Dead letter Queue AWS Lambda Client …

Slide 25

Slide 25 text

© 2021, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Confidential and Trademark Amazon Simple No=fica=on Service (SNS) Amazon SNS Fully managed publish & subscribe messaging service SNS Topic Decouple message publisher from subscribers with topics Publisher Messages pushed to subscribers Subscribers AWS Lambda Amazon SQS HTTP/S Email / SMS Amazon Kinesis Data Firehose

Slide 26

Slide 26 text

© 2021, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Confidential and Trademark Amazon Simple No=fica=on Service (SNS) Amazon SNS Fully managed publish & subscribe messaging service SNS Topic Decouple message publisher from subscribers with topics Publisher FIFO Subscribers AWS Lambda Amazon SQS FIFO HTTP/S Email / SMS Amazon Kinesis Data Firehose 1 2

Slide 27

Slide 27 text

© 2021, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Confidential and Trademark Amazon Simple No=fica=on Service (SNS) Fan Out Publisher Amazon SNS Fully managed publish & subscribe messaging service SNS Topic Decouple message publisher from subscribers with topics Subscribers AWS Lambda Amazon SQS HTTP/S Email / SMS Amazon Kinesis Data Firehose

Slide 28

Slide 28 text

© 2021, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Confidential and Trademark Amazon Simple No=fica=on Service (SNS) Publisher Amazon SNS Fully managed publish & subscribe messaging service SNS Topic Decouple message publisher from subscribers with topics Message Filtering Filter messages according to subscription filter policies Subscribers AWS Lambda Amazon SQS HTTP/S Email / SMS Amazon Kinesis Data Firehose

Slide 29

Slide 29 text

© 2021, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Confidential and Trademark Amazon Simple No=fica=on Service (SNS) Publisher Amazon SNS Fully managed publish & subscribe messaging service SNS Topic Decouple message publisher from subscribers with topics Message Filtering Filter messages according to subscription filter policies Dead-letter Queue If a subscriber is unavailable, messages can be held in a queue Subscribers AWS Lambda Amazon SQS HTTP/S Email / SMS Amazon Kinesis Data Firehose

Slide 30

Slide 30 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Common paIern with SNS & SQS: Topic-Queue Chaining Amazon SQS Queue A Backend A Publisher Amazon SNS Topic Amazon SQS Queue B Backend B Amazon SQS Queue C Backend C

Slide 31

Slide 31 text

© 2021, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Confidential and Trademark Amazon EventBridge Amazon EventBridge Fully managed event bus to connect applications Event Sources AWS Services Custom events SaaS events Events pushed to targets Rules Rules filter and send events to target Event Bus Default event bus Custom event bus SaaS event bus Decouple event sources and targets Targets (20+) AWS Lambda Amazon SQS Amazon SNS AWS Step Functions Amazon Kinesis AWS Batch Amazon API Gateway Amazon EC2 Instance Amazon ECS Task

Slide 32

Slide 32 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Amazon EventBridge Event Sources Amazon EventBridge Fully managed event bus to connect applications Event Bus Targets (20+) AWS Lambda Amazon SQS AWS Services Custom events SaaS events Default event bus Custom event bus SaaS event bus Rules Amazon SNS AWS Step Functions Amazon Kinesis AWS Batch Amazon API Gateway Amazon EC2 Instance Amazon ECS Task Dead-letter Queue If a target is unavailable, events can be held in a queue

Slide 33

Slide 33 text

Slide 34

Slide 34 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Amazon EventBridge Event Sources Amazon EventBridge Fully managed event bus to connect applications Event Bus Targets (20+) AWS Lambda AWS Services Custom events SaaS events Default event bus Custom event bus SaaS event bus Rules AWS Step Functions AWS Batch Amazon API Gateway Amazon EC2 Instance Amazon ECS Task Schema Registry Collection of event structures (schema) … Code Binding Schema Discovery Schema Publication

Slide 35

Slide 35 text

Slide 36

Slide 36 text

Slide 37

Slide 37 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark Services Comparison Amazon SQS Amazon SNS Amazon EventBridge Type Queue (Pull) Publish / Subscribe (Push) Publish / Subscribe (Push) Persistence Yes (Up to 14 days, default: 4 days) No (retry up to 23 days + DLQ) No (retry up to 24h + DLQ + Archives) Ordering FIFO FIFO No Delivery Standard: at least once FIFO: exactly once Standard: at least once FIFO: exactly once At least once Filtering Yes Yes No Consumers Millions 5 per rule / Target services Lambda (+ compute & SDK) 2 (+ HTTP/SMS/Email/Push) 20+ Throughput * Price * Standard: Almost unlimited FIFO: 3 000 TPS (soft) Standard: 30 000 TPS (soft) FIFO: 300 TPS (soft) 10 000 TPS (soft) $ per million API requests Standard ≤ 0.4 / FIFO ≤ 0.5 $ per million API requests: 0.5 $ per million events: 1 * depends on the region (us-east-1)

Slide 38

Slide 38 text

© 2021, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Confidential and Trademark Conclusion Message-Driven architectures drive resilience and elas3city è responsiveness Asynchronous Events Improve responsiveness and reduce dependencies Event Stores Buffer messages until services are available to process Event Routers Abstract producers and consumers from each other

Slide 39

Slide 39 text

Slide 40

Slide 40 text

© 2021, Amazon Web Services, Inc. or its Aﬃliates. All rights reserved. Amazon Conﬁdential and Trademark • Building event-driven applications with Amazon EventBridge • https://virtual.awsevents.com/media/1_w2mhd582 • Scalable serverless event-driven architectures with SNS, SQS & Lambda • https://virtual.awsevents.com/media/1_o5hp8egy • Decoupling serverless workloads with Amazon EventBridge • https://virtual.awsevents.com/media/1_upe6mtew • Choosing the right event-routing service for serverless: EventBridge, SNS, or SQS • https://lumigo.io/blog/choosing-the-right-event-routing-on-aws-eventbridge-sns-or-sqs

Slide 41

Slide 41 text

© 2021, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Confidential and Trademark © 2019, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Confidential and Trademark Thank you @jeromevdl