Adventures in Running a Private Cloud
summary
Comcast runs an OpenStack private cloud across 11 regions with about 37,000 VMs on roughly 2,500 hypervisors and builds tooling to manage capacity, quotas and onboarding. Monitoring uses DataDog with new caching and aggregated API requests to stop self‑DDOSes and cut a failing monitoring cycle from about 90 minutes to 12 seconds. The operation publishes SLO/SLA metrics (VM build success rates), provides quota/capacity views, and focuses on a small set of core services and customer-facing docs and bots.
AI-written summary · about
chapters (14)
- 0:00Intro
- 1:07My First Cloud
- 8:45Starting Small
- 9:0693x Growth
- 9:42Current Physical Plant
- 11:01What Do We Scale?
- 12:16Single Region
- 15:53Custom Capacity Views
- 19:58Monitoring OpenStack API Performance
- 21:24Monitoring Hypervisors (15 Iteration)
- 23:57Monitoring Hypervisors (Current Iteration)
- 26:54Support Expectations
- 30:43Scaling User Support
- 32:59Tips for Building a Private cloud
related talks
04mPaul Cleary Interview
52mLet’s build a private cloud – how hard can it be?
47mComcast Cloud Foundry Journey
23mKeynote: How Comcast is Putting Customers First - Powered by Cloud Foundry
31m10 Lessons We Learned with Cloud Foundry - Neville George, Comcast
15mSecuring the Cloud and Training Your Dragons
24mExpand Cloud Foundry for the Enterprise - Tim Leong, Comcast
24mFrom VMs To Containers: A DevOps Journey