Managing 10s of thousands in TPS in a cross-cloud with high availability.
summary
Autodesk runs tens of thousands TPS through Apigee as the north gateway while backends stay in AWS, then migrates Apigee to GCP creating an extra network hop. The team builds a multi‑region active setup with a global load balancer, uses Cloud NAT (and has to provision extra IPs) to avoid NAT port exhaustion, and monitors latency which actually drops thanks to a newer gateway version. Alternatives evaluated include cross-cloud interconnect or partner interconnect to remove public‑internet packet loss versus the lower‑cost default path.
AI-written summary · about
chapters (9)
- 0:00Autodesk and platform intro
- 3:40API gateway and traffic types
- 5:03Initial cross-cloud setup
- 8:24Migrating API Gateway to GCP
- 9:32Active-active setup challenges
- 11:22Latency and availability issues
- 13:56Availability problem causes
- 17:14Cross-cloud connectivity options
- 21:11Summary
related talks
25mAPI Management from a Network Engineer's perspective
16mHighly-Available, Global Traffic Managmement at Scale
18mTGW: Operating an Efficient and Resilient Cloud Gateway at Scale
34mLessons Learned from Five Years of Multi-Cloud
29mBuild resilient and secure global cloud networks
28mAPI Management the hybrid way
28mCloud The Final Destination
43mGetting to a Multi-Cloud Deployment - A Case Study of