How can you Scale if you Don’t Trust It?
summary
Argues that scaling requires enforceable, pervasive guard rails not meetings or manuals: a machine‑enforced policy layer that constrains CPU/memory, per‑port and per‑protocol networking, allowed runtime images and Java versions, and deployment gates tied to CI. Describes an inline proxy that issues ephemeral credentials to databases and filters or injects latency, prefers streaming access to logs (stream logs not SSH) for debugging, and evaluates Kubernetes quotas, Congress and IAM as partial but incomplete building blocks.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:29Company and product context
- 2:10Trust in production
- 4:17Policy as scalability guardrails
- 7:06Strict policy scenarios
- 9:33Pervasive policy enforcement
- 10:17Effective policy components
- 13:03Safe debugging without SSH
- 15:31Audience policy additions
- 18:41Cloud provider policy limits
- 20:25Congress and Kubernetes policy
- 21:28Policy as platform feature
- 22:10Closing and call to action
related talks
20mHow Can You Scale It If You Don't Trust It?
24mHow Can You Scale It If You Don't Trust It
38mThe Dark Side of Scale
44mScalable is Awesome, Literally!
45mPerformance-by-Design: Embedding Intelligent Scaling and Guardrails Into Platform Engineering
37mSolving Imaginary Scaling Issues, at Scale
01h 02mTools for simplified large scale systems
23mScaling Kubernetes in Regulated Environments Through Platform APIs