Architecting for Data-Intensive Workloads
summary
Argues that general purpose databases fail at high throughput and shows a concrete migration path: start relational for flexibility, add ElastiCache / MemoryDB for hot reads, move hot writes to DynamoDB + OpenSearch via Lambda for search by streaming change events, and benchmark for inflection points (example rule of thumb around 300 tps inflection with latency kicking above 50–100 ms). Presents the PI theorem: pick two among pattern flexibility, low latency and infinite scale and measures workability by TPS, latency and replication delay.
AI-written summary · about
chapters (13)
- 0:00Introduction to data-intensive workloads
- 1:48Evolution of application architectures
- 5:20Domain-driven data design shift
- 8:15Data modeling fundamentals
- 10:04Tradeoffs in database selection
- 14:21SQL vs NoSQL differences
- 17:31Scalability challenges and benchmarks
- 21:29CAP theorem for data stores
- 23:17Typical database evolution path
- 26:29Hotel booking use case overview
- 33:59Iterative data architecture design
- 36:51Applying principles to GenAI apps
- 39:00Summary and Q&A
related talks
43mArchitecting for Speed & Scale
48mDesign of data-intensive applications
43mAnalytics capabilities architectural patterns: A journey for data-driven insights
28mMoving Towards Real-Time OLAP - With First Principle Thinking and Guided by Access Patterns
55mThe Great Data Debate 2022
24mA Gentle Introduction to Building Data-Intensive Applications
28mThe Next Decade in Data Management
45mModern Data Architecture