Service Meshes: At What Cost?
summary
An open-source benchmark and playground called measuree deploys sample apps to a Kubernetes cluster and runs load tests to quantify cost: three functional adapters today (Istio, Linkerd, Consul), over ten meshes profiled overall, and tests run in a single-node Ubuntu Kubernetes VM (24 GB RAM, 8 cores). Results show per-sidecar CPU and memory vary widely, control-plane overheads differ, and telemetry/feature choices materially increase cost; methodology includes university collaboration and PhD research.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:15Service Meshes
- 2:25Challenges
- 4:13Definitions
- 7:00Deployment Models
- 9:00Open Source Project
- 10:15Architecture Mastery
- 11:15Service Mesh Interface
- 14:55Girish Ranganathan
- 23:15Memory
related talks
33mService Meshes: Istio, Linkerd - or No Mesh at All?
49mSolving the Service Mesh Adopter's Dilemma
29mYou Probably DON’T Need a Service Mesh!
09mLightning Talk: Dissecting Service Mesh Overheads
44mService Mesh - The New Single Point of Failure
34mService Mess to Service Mesh
40mComparison of Current Service Mesh Architectures
43mWTF Do We Need a Service Mesh?