Orchestrating thousands of bots from the cloud
summary
Claims a cloud migration of real-time orchestration that preserves millisecond predictability and throughput: uses AWS Outposts as an insurance policy for remote sites, adds Keyspaces and Network Load Balancers to hit a target of <50 ms predictable messaging delay, runs simulations and staged trials (chill then ambient grids), migrates all orchestration in 12 months under Project Tempest, moves the largest site first, orchestrates 5,000 then nearly 10,000 bots, and reports no significant outages.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 0:57Our history
- 1:31Cloud first approach
- 1:49Automated warehouses
- 2:47Online orchestration
- 3:41Orchestrating a fleet
- 4:49Cost
- 5:56AWS Outpost
- 6:30Tradeoffs
- 7:24Team effort
- 8:06Cloud migration
- 9:11Stretching goals
- 9:38Whats next
- 10:21Outro
related talks
25mMass migrations to the cloud
41mA Future Journey: How to Migrate 100 Clusters Between Clouds Without Downtime?
17mDeployments We Can Believe In
34mScaling and Orchestrating “Good Bot” With Kubernetes
16mFrom Complexity to Clarity
20mMigration of 300 microservices from AWS Cloud to
01h 08mPerformance testing in the cloud
01h 03mContinuous Security in the Cloud