Duplicating data or replicating data in Micro Services
13 likes7,485 views
The document discusses the differences between monolithic and microservices architectures, emphasizing the need for loosely coupled systems in complex applications. It highlights the importance of defining service boundaries, avoiding data duplication across services, and utilizing replication for performance optimization. Additionally, it presents the benefits of messaging and event-driven architecture for effective communication between services.
Duplicating data or replicating data in Micro Services
- 1. in a Microservices Architecture Dennis van der Stelt Software Architect
- 2. v a n d e r s t e l t 2 Your presenter Dennis van der Stelt http://dennis.bloggingabout.net/ Software Architect Geek Developer Architect Gamer Enthousiast Professional Skills Something about me that you should know About Dennis Innovative software architect Almost 20 years of experience in development of distributed systems. Has a continuous drive to learn and improve knowledge in different architectural styles, including quality in software development. Highly motivated to share his knowledge via articles, presentations and his blog. Awards 98% 80% 60% 40% 100% NServiceBus Champ Typemock MVP Certified Trainer
- 3. v a n d e r s t e l t 3 Agenda In a polyglot architectural landscape with a lot of business logic, data and information, how can we build solid loosely coupled systems? Let’s have a look at the differences between those opposites Monolithic vs Microservices Dividing your system’s business logic over logical boundaries Service Boundaries Why duplicating data is a bad practice, but replicating data is perfectly fine Replicating vs Duplicating
- 4. v a n d e r s t e l t Monolithic application Wikipedia : large, powerful, and intractably indivisible and uniform. A monolithic application is build and deployed as a single unit. WordPress is a successful & scalable monolithic application used by millions! Monolith is therefor not by default a badly designed system 4
- 5. v a n d e r s t e l t 5 Microservices a software architectural style for complex applications of small, independent processes. services are small, highly decoupled and focus on a single task communicate via language agnostic APIs boundaries are explicit services are autonomous share schema & contract, not classes compatibility is based on policy
- 6. v a n d e r s t e l t Monolithic application 6 single deployment unit
- 7. v a n d e r s t e l t Microservices 7 multiple deployable services
- 8. v a n d e r s t e l t Authentication 8 Coupling OSS ViewModel Change Product Database Razor
- 9. v a n d e r s t e l t 9 First law “My First Law of Distributed Object Design: Don't distribute your objects” Martin Fowler in Patterns of Enterprise Application Architecture
- 10. v a n d e r s t e l t 10 Monolithic application problems Problem Analysis of all monolithic systems are build with high time constraints, resulting in big balls of mud in which no developer can find their way. 42% Maintenance The coupling makes it hard to change anything Documentation Cost to add new features Building the 100th feature as fast as the 1st Project Ownership
- 11. v a n d e r s t e l t Monolithic application 11
- 12. v a n d e r s t e l t 12
- 13. v a n d e r s t e l t 13 Microservices “don't even consider microservices unless you have a system that's too complex to manage as a monolith” Martin Fowler
- 14. SERVICE BOUNDARIESDividing your system’s business logic over logical boundaries
- 15. v a n d e r s t e l t 15 Microservices architecture Deployment Don’t necessarily need to separately deploy every single service Services boundaries Define boundaries around logical business processes Monitoring Document message flows and monitor them with great tools Messaging Asynchronous communication via messaging & publish/subscribe
- 16. v a n d e r s t e l t Any organization that designs a system will produce a design whose structure is a copy of the organization's communication structure. “ ”Conway’s law – 1968
- 17. v a n d e r s t e l t Service boundaries 17 Shipping FinanceSales Marketing
- 18. v a n d e r s t e l t Deployment 18 Shipping FinanceSales Marketing SQL SQL { }
- 19. v a n d e r s t e l t Architectural style 19 Shipping FinanceSales Marketing CQRS Event Sourcing CRUD there is no top level architecture
- 20. v a n d e r s t e l t Coupling aspects 20 SpatialTemporalPlatform
- 21. v a n d e r s t e l t Messaging 21 Shipping FinanceSales Marketing SQL
- 22. v a n d e r s t e l t Messaging 22 Shipping FinanceSales Marketing SQL
- 23. v a n d e r s t e l t Messaging 23 Shipping FinanceSales Marketing SQL Publish / Subscribe
- 24. v a n d e r s t e l t 24 Messaging Handle business events through Inversion of communication by supplementing SOA with EDA “ ”
- 25. v a n d e r s t e l t 25 Resume Service Boundaries Try to see if your organizational communication structure can be the model for your logical service boundaries. Messaging Avoid RPC style communication between services. Use messaging with an event driven architecture. Deployment Don’t necessarily see every microservice as a deployable unit. Do keep an eye out for coupling. Avoid joins in your database between entities.
- 26. DUPLICATION vs REPLICATIONWhy you shouldn’t duplicate, but replicate data
- 27. v a n d e r s t e l t 27 Do not duplicate across service boundaries You can share… some information Publish / Subscribe to notify in advance
- 28. v a n d e r s t e l t 28 Duplicating data What event just took place business process related events that other services require to properly function and take action on Status changes Identifiers Date Time Info Who was the subject the triggered event has a specific subject in our domain When does the status expire when status updates expire or are no longer relevant at a point in time, we provide expiration information
- 29. v a n d e r s t e l t Sales Messaging 29 Taxi Customers TaxiRequested TaxiRequested RequestTaxi
- 30. v a n d e r s t e l t Sales Messaging 30 Taxi TaxiRequested RequestTaxi Finance TaxiAssigned Customers
- 31. v a n d e r s t e l t Sales Messaging 31 Taxi RequestCancelled Finance RequestCancelled CancelRequest
- 32. v a n d e r s t e l t 32 Replicating vs Duplicating data So duplication of data is not good But what about this replicating data?
- 33. v a n d e r s t e l t 33 Replicating vs Duplicating data So duplication of data is not good But what about this replicating data? Replicating within a service boundary is okay for performance or connectivity issues
- 34. v a n d e r s t e l t 34 Replicating data options Performance optimization Suitable for all categories business and personal presentation In memory cache Prepare for optimal querying Prepare data for querying, for example into a specific readmodel / viewmodel Resolve latency issues Replicating data across physical machines does not break service boundaries Resolve network issues Different sites across the globe, or a moving site without connection, like a large container ship Viewmodels Different machines Different sites
- 35. v a n d e r s t e l t 35 Composite UI Sales Lists requests made by customer Customer Displays customer information such as name and current location Taxi Shows where taxi drivers are and at what time they arrive at your location Finance Displays status of all financial transactions and outstanding bills
- 37. v a n d e r s t e l t 37 Monitoring
- 38. v a n d e r s t e l t 38 Monitoring
- 39. v a n d e r s t e l t 39 Monitoring
- 40. v a n d e r s t e l t 40 Resume; how we build microservices Define Boundaries Define boundaries of your services. Your organizational structure may help. Do not duplicate data across services. Share status changes, identifiers and date-times. Implement Implement your microservices that handle a single task. Use publish/subscribe to notify other services with status changes. Composite UI Everything from data up until the user interface should belong in a service, resulting in a composite ui that retrieves data based on the mentioned identifiers. Deploy Deploy your microservices, but not necessarily as single units. A possibility is a deployment unit per service boundary.
- 41. v a n d e r s t e l t 41 Microservices can be hard to build, deploy & monitor. Get help from people with experience Think about temporal coupling between services and how to solve this, possibly with messaging Do not duplicate data between services, as at some point business logic will follow Conclusion
- 42. v a n d e r s t e l t 42 Contact Me Address Rotterdam, The Netherlands Email dennis@bloggingabout.net Twitter @dvdstelt Website http://dennis.bloggingabout.net Thank you for your time…