How We Built a Framework at Twitter to Solve Service Ownership & Improve Infrastructure Utilization
summary
Twitter builds a production system that models, meters and charges infrastructure across Aurora/Mesos, Hadoop and Manhattan, then ties usage to teams so accountability drives optimization. The team ships Kite — a unified service manager and a chargeback metering pipeline that ingests scheduler and cluster metrics, resolves owners, computes time-varying unit costs from bare‑metal TCO, and exposes bills and quota. It already aggregates 10,000 identifiers and 1,000 projects and reduces unknown owners and wasted capacity.
AI-written summary · about
chapters (9)
- 0:00Intro
- 0:13Agenda
- 1:12History Context
- 4:16Chargeback
- 13:08Reporting
- 15:54Learnings
- 17:52Introducing KITE
- 21:08Impact of KITE
- 24:56Resources for future work
related talks
06mLightning Talk - Micro-Services Lifecycle Management at Twitter
51mHow We Built a Metering & Chargeback System to Incentivize Higher Resource Utilization
23m"In the Belly of the Whale: Operations at Twitter"
31mScaleable Service Frameworks
26mRise of the Machines Security Automation at Twitter
42mHow Houghton Mifflin Harcourt Went from Months to Minutes for Infrastructure Delivery
21mTeam Ownership - Who Owns Production
54mDevOps in Real Life, A How-To