A service mesh costs 0.16ms at one connection and 86% of your throughput at 32

8.2 relevance

Rigorous performance benchmark with actionable findings for service mesh users.

DevTools dev.to

Answer the board on AI readiness

Arrive with evidence, not opinion. See how your foundation benchmarks against your peers, get a clear read, and leave with a ranked place to start.

Why It Matters

For engineers evaluating service meshes like Linkerd for production, this benchmark reveals that standard latency figures are dangerously misleading, as the proxy's CPU overhead causes throughput to collapse under realistic concurrency, directly impacting capacity planning and infrastructure costs.

Key Takeaways

Benchmark your service mesh under realistic concurrency (32+ connections) rather than single-connection latency, as proxy CPU overhead can collapse throughput by over 80% even with idle hardware.

A service mesh costs 0.16ms at one connection and 86% of your throughput at 32

Summary

A benchmark of Linkerd on DigitalOcean Kubernetes reveals a single-connection latency cost of just 0.16ms, but at 32 concurrent connections throughput collapses by 86% and latency jumps 3.2ms. The sidecar consumed 2-3x the CPU of the nginx application container, with node utilization only hitting 50%, proving the proxy software itself is the bottleneck, not the hardware. This exposes how standard single-connection benchmarks mask severe throughput degradation under realistic concurrency.

Author

Remdore