How Netflix Manages $18B of Content Spend With Brian Orth and David Johnson | Current 2022
Netflix spends an ever increasing amount on content to bring joy to our members – over 18 billion dollars in 2022! As a business, how does Netflix ensure that our forecasted spend is accurate? How do we enable systems and business processes to be able to move in a highly aligned, loosely coupled way that is so critical to the Netflix Culture? Come learn how Netflix Content Controllership Engineering uses Kafka to provide a highly available, ordered stream of data that allows our fleet of financial services to have a measurable increase in forecasting model accuracy by using data from Actual Spend. All while taking advantage of Netflix Engineering investment into our Studio API. We’ll go into the details of building observability into our event-driven architecture, talk through event data enrichment, and explain how we are building scalability and flexibility into our design.
3. 3
Q: How many countries did
Stranger Things Season 4
become #1 in the
week after launch?
Break The Ice
4. 4
Building largest Studio in the World
Roughly $18b in Content Spending
Netflix Animation Growing Rapidly
Expanding to Netflix Games!
Netflix Studio Scaling
13. 13
● Flexible design to enable accounting workflows
● Traceability
● Lightweight
● Single Writer
Accounting
System
Triggers Export Of
Raw Events
Event Routing Service
Event Routing Service
14. 14
● Horizontally scalable
● Parallelizable
● GraphQL source of truth
Enrichment
Raw
Events
Enriched
Events
Enrichment
23. 23
● Alerting – report when something goes awry
● Discovery – determine what isn’t working
● Diagnosis – debug why something isn’t working
Observability
24. 24
Domain Challenges
● Translating data model between
producers and consumers
● Adding supplementary data
Technical Challenges
● Ordering / idempotence
● Failure cases
● At least once
Challenges
25. 25
● Able to answer business questions in near real-time
● Automated user workflows
● Consumer architecture decoupled
Results
26. 26
● Event Trigger from Application
○ Future Change Data Capture
● In the future Kafka can be source of truth
with log compaction
● Additional Consumers without needing to
change the architecture
● Streaming as the backbone of the Studio
Finance Connected Data
Coming Soon