Service Mesh - The New Single Point of Failure
summary
Comparison of concrete choices and tools: Istio consolidates complexity into a monolithic control plane and removes Mixer by moving policy into Envoy via WebAssembly, and uses revisions-based upgrades (install a parallel control plane and cut namespaces over) plus analyzers and optional Prometheus integration. Linkerd ships a pre/post-install check command, a live-request tap command, and relies on fail‑open defaults and multi‑cluster secrets for federation. Consul offers an RBAC bridge, a Prometheus plugin for topology metrics, ACL automation, and a single installation-per-cluster safety decision.
AI-written summary · about
chapters (35)
- 0:00Intro
- 0:21About Sabeen
- 0:56About Mitch
- 1:46About Thomas
- 2:50Complexity
- 3:28Service Mesh
- 5:34Microservices
- 8:40Prometheus Integration
- 10:19Mixer Architecture
- 12:05Multiple Console Installations
- 13:32Multiple Cluster Installations
- 14:35Single Point of Failure
- 16:16Upgrades
- 17:06Managing Complexity
- 17:35Automated ACL Setup
- 18:55RBAC Bridge
- 21:19Analyzers
- 22:57Events
- 23:52Federation
- 24:53Its not done
- 26:12The unsung heroes
- 27:29Enterprise grade software use case
- 28:12A difficult problem
- 28:56An anecdote
- 29:56A matter of maturity
- 31:36How to dump service discovery information
- 33:09Proactive Alerts
- 33:39Mesh Debugging
- 36:58Remote Development
- 38:04Failure Static
- 39:25Design for Failure
- 40:27Data Plane vs Control Plane
- 41:56Wrap Up
- 42:26Getting Involved
- 43:23Outro
related talks
30mService Mesh Is Still Hard. The Things we all Did Wrong & Tales of Woe
33mService Meshes: Istio, Linkerd - or No Mesh at All?
34mService Mess to Service Mesh
40mReimagining Application Networking and Security
26mService Mesh Journey - Building Security and Reliability One Si...
52mService Mess to Service Mesh
28mService Meshes: At What Cost?
53mDo you need a Service Mesh?