Drive positive impact with feature flags and data (PRT053)
summary
Migrating pipelines from Amazon Kinesis to Kafka and from S3 to MongoDB using traffic mirroring and tap compare testing ensures zero downtime while targeting zero lost events, no latency increase, and no throughput decrease. Mirrored requests are fire-and-forget via try/catch, Kafka/S3 replicas are discarded by the live path, and per-request latencies and ops/sec are posted to Split to create metrics and causal analysis that attribute changes to the migration. Customer survey numbers: 15% some flags, 23% most, 8% bringing data, 50%+ experimenting.
AI-written summary · about
chapters (5)
- 0:00Start
- 1:27Moving towards an experimentation mindset
- 4:04What is Traffic Mirroring?
- 4:28Amazon Kinesis to Kafka Migration
- 7:06Amazon S3 to MongoDB Migration
related talks
21mMigrate Data, Mesh in mind
19mConfident Data Migration: automatic regression testing
13mData for product
27mHow to be data-driven when you aren’t Netflix (or even if you are)
31mOne Data Pipeline to Rule Them All
32mMove to managed analytics
28mMigration of a realtime stats product from Storm to Flink
27mTinkertoys, microservices, and feature management...