Service Mesh Big Survey
summary
Survey maps service meshes against five classification axes — control plane, data plane, API, traffic interception and extensions — and shows concrete tradeoffs: sidecars cost CPU but limit blast radius, node‑level proxies save resources but expand attack surface, proxy‑less (gRPC) cuts latency at the cost of flexibility. Demoed a Lazy xDS optimization that sends only needed configs to pods, reports Lazy xDS in use, and cites 200 production installs since 2018 with tens of thousands of pods and 17 support teams.
AI-written summary · about
chapters (23)
- 0:00Introduction
- 1:13What is service mesh
- 3:04Service mesh components
- 5:54Control Plane architectures
- 7:17Data Plane implementations
- 10:06Traffic interception
- 12:56Service mesh classification
- 13:38Istio overview
- 15:48Linkerd overview
- 16:50Consul overview
- 17:35Kuma overview
- 18:18NGINX Service Mesh overview
- 19:23Cilium overview
- 20:46Open Service Mesh overview
- 23:59Traefik Mesh overview
- 24:42Synapse Service Mesh overview
- 26:10Scaling challenges
- 27:34Cluster scaling
- 28:59Security impact
- 31:05Operational complexity
- 32:29Community and standards role
- 33:55Summary
- 34:58Q&A
related talks
32mService Mesh - Beyond the Hype
31mClean-up Your Service Mess with Service Mesh
37mThe Service Mesh Wars: A New Hope for Kubernetes
50mDistributed systems patterns with Spring Cloud, Service Meshes and eBPF
41mService Mesh beyond single clusters
48mProxies, Gateways, and Service Mesh. Why Are Words so Confusing?
51mReplicated service mesh: hardening systems against failure modes in load balancing
43mService MESH without the MESS