EDD - UBC - June 2020
0 likes65 views
Sasha is the founder of Persistr, an event-driven design framework and platform. He coined the term "event-driven design" in 2019 to describe building software where data is versioned with semantic domain events that represent changes in state. Event-driven design allows reconstructing any domain object's state by replaying its event stream and caching projections in databases for querying.
EDD - UBC - June 2020
- 2. PERSISTR About Sasha Founder of Persistr: an EDD framework and SaaS platform. Previously, Sasha was Director of Engineering at Kater, a Mobility-as-a-Service company, where he built an EDD rideshare app to compete with Uber and Lyft. Currently, VP Software Dev. Engineering at Mastercard Persistr is an Open Source client library and server software that allows developers to quickly and easily build backend and mobile applications following Event-Driven Design principles Sasha Jolich L: www.linkedin.com/in/sasadjolic P: +1-604-998-2260
- 3. PERSISTR About Sasha Inspired by MIT AI Labs subsumption architecture Event-driven game code at EA Sports (Vancouver) Practicing event-driven architectures since 1997 Met Greg Young in 2008 (eventsourcing) UBC with Adam Dymitruk (event modelling)
- 4. PERSISTR About Sasha Practicing event-driven architectures since 1997 Most recently, built a ridesharing app called Kater following EDD principles
- 5. PERSISTR About Sasha Event-Driven Design (EDD) is a term I coined in summer of 2019
- 6. PERSISTR About Sasha Event-Driven Design (EDD) is a term I coined in summer of 2019 To describe the differences between EDD and other patterns used in the industry such as event sourcing, event streaming, CQRS, etc
- 7. PERSISTR About Sasha Event-Driven Design (EDD) at OReilly Software Architecture NYC in Feb 2019
- 8. PERSISTR About Sasha COMING SOON Patterns of Event-Driven Design 息 2020 A practical guide to building event-driven applications
- 10. PERSISTR #EDD Event-Driven Design is a way of building software where data is versioned with semantic events
- 11. PERSISTR #EDD Semantic Events are meaningful
- 12. PERSISTR #EDD They are domain events
- 13. PERSISTR #EDD Domain events are facts about what happened in the domain Customer booked ride Customer picked up Customer arrived at destination Customer paid bill & tipped driver Driver available Driver assigned Driver accepted Trip started Trip completed Trip booked Customer booked ride Driver available Trip booked CustomersDrivers Trips
- 14. PERSISTR #EDD Each domain event represents a change in state of a domain object Customer booked ride Customer picked up Customer arrived at destination Customer paid bill & tipped driver Driver available Driver assigned Driver accepted Trip started Trip completed Trip booked Customer booked ride Driver available Trip booked CustomersDrivers Trips { customer: John Doe, pickup: 1234 Granville St, dropoff: 321 Cambie St }
- 15. PERSISTR #EDD Every domain object has an event stream containing its entire change history Driver Customer Trip
- 16. PERSISTR #EDD Rebuild any domain objects state by replaying its event stream Driver Customer Trip
- 17. PERSISTR Driver #EDD Cache state of domain objects in a traditional database Customer Trip SQL or NOSQL
- 18. PERSISTR Driver #EDD Cache state of domain objects in a traditional database Customer Trip SQL or NOSQL This is called a projection
- 19. PERSISTR Driver #EDD Cache state of domain objects in a traditional database Customer Trip SQL or NOSQL Projections are lossy and one-way
- 20. PERSISTR Driver #EDD Cache state of domain objects in a traditional database Customer Trip SQL or NOSQL Event streams + Projections > Traditional
- 21. PERSISTR What are some interesting applications of EDD?
- 22. PERSISTR EDD Patterns Documentation-as-code with executable BDD scenarios Given pattern of past events When changing state Then expect new event Customer booked ride Customer picked up Customer arrived at destination Arrive at destination
- 23. PERSISTR EDD Patterns Detect event patterns Customer picked up Driver accepted Trip started Customer arrived at destination Customer paid bill & tipped driver Driver assigned Trip completed Driver available Trip booked Customer booked ride
- 24. PERSISTR EDD Patterns Detect event patterns, capture metrics, and take action Customer picked up Driver accepted Trip started Customer booked ride 30 secs 2 mins 15 secs Update drivers ranking based on performance
- 25. PERSISTR EDD Patterns Secure access to events and streams with roles and permissions Driver Customer Picked up Arrived at destination Paid bill & tipped driver Assigned Accepted Booked ride Available Read allowed events Read allowed streams Append to streams Delete events/streams
- 26. PERSISTR EDD Patterns Customer signs up and verifies their email. Customer then books a new trip. A few moments later, customer is notified of a driver that will pick them up. Coordinate between components Verifications service verifies customer email Bookings service books new customer trips Assignments service assigns drivers to trips
- 27. PERSISTR EDD Patterns Coordinate between components Customer signs up and verifies their email. Customer then books a new trip. A few moments later, customer is notified of a driver that will pick them up. Driver accepted Driver assigned Customer picked up Trip started Customer arrived at destination Customer paid bill & tipped driver Trip completed Driver available Trip booked Customer booked ride Managed by Assignments service
- 28. PERSISTR EDD Patterns Coordinate between components Customer signs up and verifies their email. Customer then books a new trip. A few moments later, customer is notified of a driver that will pick them up. Driver accepted Driver assigned Customer picked up Trip started Customer arrived at destination Customer paid bill & tipped driver Trip completed Driver available Trip booked Customer booked ride Managed by Bookings service
- 29. PERSISTR EDD Patterns Customer names, credit card numbers, birth dates Trip history, current location, frequently visited locations Scrub sensitive PII and/or PCI data from the shadow production-like environment Production Shadow SCRUB & COPY
- 30. PERSISTR EDD Patterns Distribute data across multiple geographic regions North America Europe MIRROR Australia South America
- 31. PERSISTR EDD Patterns Decentralize data across multiple geographic regions North America Europe Australia South America
- 32. PERSISTR What is the Computer Science behind EDD?
- 33. PERSISTR Computer Science State = Object state given objects event history Domain Objects = State + Behaviour Behaviour = Finite State Machine + Time + Reactions Customer
- 34. PERSISTR Computer Science Encode behaviour as a finite state machine Make behaviour data driven React to state machine transitions or state changes
- 35. PERSISTR What kind of innovation does EDD enable?
- 37. PERSISTR Innovation EDD makes practical nanoservices possible Domain objects reside in different regions
- 38. PERSISTR Innovation EDD makes practical nanoservices possible And can relocate from one region to another
- 39. PERSISTR Innovation EDD makes practical nanoservices possible Domain objects can be mirrored in other regions in case of network disconnects
- 40. PERSISTR Distributed, decentralized, fault-tolerant, language-agnostic network of Domain Objects Innovation EDD makes practical nanoservices possible