Building Personalization with Orleans and Actor Modelling

Explore actor modeling for personalization using Orleans in this conference talk. Learn why Orleans, an open-source Virtual Actor platform, was chosen to build a new personalization system for favorites and progress at tv.nrk.no and radio.nrk.no. Dive into the differences of actor modeling, examine key design decisions, and see how it was implemented using F#. Discover the importance of build pipelines and deployment strategies for Orleans applications, including Docker builds, Azure Kubernetes Service deployment, and monitoring with Application Insights. Gain insights into actor behavior, state injection, network latency, catalog management, favorites handling, position tracking, state storage, version control, pattern matching, and Etag caching. Understand the lessons learned about idempotency, deadlock prevention, logging, and deploying to production environments.

Introduction
Meet Harald
Agenda
Relevance
Dynamic content
Big ball of mud
Personalization
Actor Systems
Orleans Runtime
Actor behaviour
Actor state injection
Actor proxy
Network latency
Actor model
Catalog
Favorites
Positions
Favorite
Bloom Filter
State Storage
Version State
Pattern Matching
Etag Caching
Reporting
Operations
Docker images
Web API
Silo hosts
Preproduction
Fresh State
Dashboard
Grains
Single grains
Application Insights
Lessons Learned
Idempotency
deadlock
position grains
logging
direct clients
deploy to production
storage