Building the future of time series data
summary
Uses Cosmos DB (Cassandra API) as the storage for a global IoT pipeline handling ~5,000 edge devices and peaking at 2 million observations/min, messages 8–10 KB with 10–50 observations each. Data model uses primary key (sensorId, observation, timestamp) with timestamp as a descending clustering key, which forces supplying observation names for many reads; workaround persists observation lists in Azure Table. Writes show p95 <10 ms; operational changes include move RUs to keyspace, manual vs. autoscale tradeoffs, and archiving history to blob.
AI-written summary · about
chapters (6)
related talks
10mBuilding IoT Apps with JSONB & Timeseries
49mThinking in time series, the new solution to your pro...
30mScaling Time-Series Data to Infinity: A Kubernetes-Powered Solution with Envoy
39mA New Data Model for Time Series Data Processing- 39mA New Data Model for Time Series Data Processing
21mBuild Scalable Inventory System with Azure Cosmos DB
01h 03mTime Traveling in the Cloud
49mIoTDB - Moving from Relational to Timeseries