Migrating Monolithic Applications with the Strangler Pattern
3 likes1,446 views
This document discusses strategies for migrating monolithic applications to the cloud using the strangler pattern. It begins with an overview of the strangler pattern, which involves gradually building a new system around the edges of an existing monolith. It then provides examples of how to implement the strangler pattern on AWS by hosting the existing application, adding facades with API Gateway, detecting hot spots with X-Ray, replacing hot spots with Lambda functions, and iteratively strangulating more of the monolith over time until it is retired. The document emphasizes that this incremental approach allows migrating applications at a lower cost and risk compared to full rewrites.
Migrating Monolithic Applications with the Strangler Pattern
- 1. 1 Migrating Monolithic Applications with the Strangler Pattern Thanh Nguyen Senior Solutions Architect BUILDING MODERN APPLICATIONS 🌥🎯🚀
- 2. 2 q Strangler Pattern 101 q Strangler Pattern Strategy in action q Data Access Patterns q Canary release q Deploy & Scale Monolith using ElasticBeanstalk PaaS Agenda
- 3. 3 Reasons to Modernize Cost reduction Data center consolidation Digital transformation Agility/dev productivity Business Modernization
- 4. 4 The 6 Rs: Application Migration Strategies R e - a r c h i t e c t ( r e w r i t i n g / d e c o u p l i n g a p p l i c a t i o n s ) Redesign application/ infrastructure architecture Application code development Full ALM/ SDLC Integration Re-purchase (replace – drop & shop) Purchase COTS/ SaaS & licensing Manual install & setup Re-platform (lift & reshape) Modify underlying infrastructure Determine new platform Retain/ revisit Manual Manual config Manual deploy Manual install Retire/ decommission Re-host (lift & shift) Discover/assess/prioritize applications Determine migration path Use migration tools Transition Production Validation Automate
- 5. 5 Re-host migration strategy The re-host migration strategy, referred to as the “lift and shift” approach, is to move applications without changes In large-scale, legacy migrations, organizations are looking to move quickly to meet business objectives The majority of these applications are re-hosted Discover/assess/ prioritize applications Determine migration path Automate Use migration tools Validation Transition Production
- 6. 6 Comparison of Migration Strategies Effort (time & cost) Opportunity to optimize Retire N/A N/A Retain N/A Re-host Re-purchase Re-platform Re-architect Strangler
- 7. 7 Martin Fowler’s Strangler Pattern “…gradually create a new system around the edges of the old, letting it grow slowly over several years until the old system is strangled.” Martin Fowler June 29, 2004
- 8. 8 Creating a new system around the edges Monolith Browser DB Proxy Browser Service DB Proxy Browser Service DB Service DB Service DB DB DB Monolith Monolith
- 10. 10 Corporate Data Center Monolith AWS Cloud VPC Availability Zone 1 Availability Zone 2 Web & App layer Amazon EC2 instances ELB Amazon RDS master Database layer Synchronous replication RDS standby Region Users Auto Scaling Group EC2 instances Step 1: Re-host application
- 11. 11 Corporate Data Center Monolith AWS Cloud VPC Availability Zone 1 Availability Zone 2 Synchronous replication RDS standby Region Users EC2 instances Step 1: Re-host application Amazon EC2 instances ELB Amazon RDS master Auto Scaling group Web & App layer Database layer
- 12. 12 Step 2: Facade with Amazon API Gateway AWS Cloud VPC Availability Zone 1 Availability Zone 2 Region Users API Gateway EC2 instances ELB Amazon RDS master Synchronous replication RDS standby Auto Scaling group EC2 instances Web & App layer Database layer
- 13. 13 Step 2: Facade with API Gateway – Setup
- 14. 14 AWS Cloud VPC Availability Zone 1 Availability Zone 2 Region Users API Gateway Amazon CloudWatch EC2 instances ELB Amazon RDS master Synchronous replication RDS standby Auto Scaling group EC2 instances Step 3: Detect & measure hot spots Web & App layer Database layer
- 15. 15 Step 3: Detect & measure hot spots AWS Cloud VPC Availability Zone 1 Availability Zone 2 Region Users API Gateway AWS X-Ray EC2 instances ELB Amazon RDS master Synchronous replication RDS standby Auto Scaling group EC2 instances Web & App layer Database layer
- 16. 16 AWS Cloud VPC Region Users API Gateway Lambda function hot spot 1 EC2 instances ELB Amazon RDS master Synchronous replication RDS standby Auto Scaling group EC2 instances AWS Database Migration Service Amazon DynamoDB Step 4: Create AWS Lambda for top hot spot Web & App layer Database layer
- 17. 17 Write Read Synchronize Use case(s) • Data structure stays consistent • Data content stays consistent • Eventual consistency acceptable o Batch synchronize • Golden copy remains with old database Old New Monolith Strangler Synchronization technology(s) • Triggers • AWS Database Migration Service • Generic batch process Read L-shaped pattern
- 18. 18 Write Read Synchronize Use case(s) • Data structure differs • Data content is expanded upon • Eventual consistency acceptable o Batch synchronize • Golden copy for expanded schema to new database and remains with existing data in the old database Old New Monolith Strangler Synchronization technology(s) • Events • AWS Database Migration Services • Generic batch process Read Backwards L-shaped pattern
- 19. 19 Write Read S y n c h r o n i z e Old New Monolith Strangler Read S y n c h r o n i z e Use case(s) • Data structure differs • Data content is expanded upon • Immediate consistency required • Golden copy moves for the expanded data to the new database and remains with existing data in the old database Synchronization technology(s) • Amazon SQS • Amazon Kinesis • Amazon MQ • Amazon MSK • AWS Lambda • Generic queue process Inverted Y-shaped pattern
- 20. 20 Write Read Use case(s) • Data structure remains consistent • Data content remains consistent • Immediate consistency required • Golden copy remains with old, and only, database Old Monolith Strangler Synchronize technology(s) • Amazon ElastiCache • Generic caching Read Slanted V-shaped pattern
- 21. 21 Broad Database and Analytics Services Portfolio Relational Databases Nonrelational Databases Data Warehouses Hadoop and Spark Amazon Redshift Amazon EMR Operational Analytics Amazon Elasticsearch Service Amazon Aurora Amazon DynamoDB Business Intelligence Amazon QuickSight Amazon RDS Amazon DocumentDB Amazon ElastiCache Real-time Analytics Amazon Managed Streaming for Apache Kafka PostgreSQL Logstash Elasticsearch Kibana Microsoft SQL Server Microsoft SQL Server
- 22. 22 AWS Cloud VPC Region Users API Gateway Lambda function hot spot 1 EC2 instances ELB Amazon RDS master Synchronous replication RDS standby Auto Scaling group EC2 instances Amazon DynamoDB AWS Database Migration Service Function invocations T h e i m a g e p a r t w i t h r e l a t i o n s h i p I D r I d 2 8 w a s n o t f Step 5: Canary release Web and app layer Database layer
- 23. 24 X-Ray canary service map
- 24. 25 Step 6: Iteratively strangulate AWS Cloud VPC EC2 instance EC2 instance ELB Amazon RDS master Synchronous replication RDS standby Region Users Auto Scaling group API Gateway Lambda function hot spot 1 Function invocations DynamoDB EC2 instances ELB Amazon Aurora AWS Database Migration Service Web and app layer Database layer
- 25. 26 Step 7: “Not with a bang but a whimper” AWS Cloud VPC ELB Region Users API Gateway Lambda function hot spot 1 DynamoDB EC2 instances ELB Aurora RDS AWS Fargate Function invocations
- 26. 27 Step 7: ”Not with a bang but a whimper” AWS Cloud VPC Region Users API Gateway ELB API Gateway Lambda function hot spot 1 Function invocations DynamoDB EC2 instances ELB Aurora RDS Fargate
- 27. 28 Takeaways • “Just move it”1 (Jim Scholefield, Nike Global CIO) is often good advice • Customers have seen 25%-30% saves without going “cloud native” • Re-hosting offers lowest time to value and is a platform for more value • Re-architecting might be appropriate for applications with a bright future • Consider the strangler pattern to realize that value incrementally • AWS services like API Gateway and Lambda are accelerators • And, they don’t require you to “spend to save” – no extra infrastructure • Leverage canaries and other modern techniques to validate new features