How We Built a Chargeback System for Twitter’s Compute Platform (Mesos/Aurora)
summary
Built a centralized Cloud Platform Management System that catalogs, plans, provisions, meters and chargebacks multi-tenant compute across Mesos/Aurora and other platforms, tying usage to service identity so platform services can bill in real dollars. Implementation uses containers for relocatable workloads, Finagle for RPC patterns, LDAP for human auth and recommends Spiffe-style service tokens for service-to-service auth. The system surfaces real costs and drives measurable improvement in compute utilization after onboarding.
AI-written summary · about
chapters (10)
- 0:00Start
- 8:04PRIMARY OBJECTIVES BY FUNCTION
- 9:05A FINE GRAINED SERVICE MODEL
- 13:58A FULL LAYERED APPROACH
- 15:04LAYERS AND VARIATIONS
- 16:05MULTI-TENANT NEIGHBORHOODS
- 19:29WHAT IS SECURITY?
- 20:46SERVICE IDENTITY
- 24:04CONTAINER BASED MULTI-TEMANT SERVICES
- 26:02KEY TAKEAWAYS
related talks
27mHow to get rich with cloud!
48mCloud FinOps & Kubernetes Optimisation at Scale
19mKeynote: Operating 600+ Mesos Servers on 7 Datacenters @Criteo
45mMesos, Then and Now
38mFinancial Resiliency Engineering: Taming Cloud Costs
24mPlatform Engineering Is the Key Lever for Sustainable Cloud Infrastructure
26mTwo sides of the platform coin with operator's and customer's voices
49mPublic cloud in regulated industry, not just swipe of a cred