Is Spring Cloud Gateway Really Slower Than Zuul? Benchmark Results Reveal Surprising Speed

Motivation

Many developers repeat the claim that Spring Cloud Gateway performs worse than Zuul, questioning why Spring would develop its own gateway if it were truly inferior.

Background

The claim originates from a benchmark that used ab (ApacheBench) to measure throughput, which showed Spring Cloud Gateway lagging behind Zuul.

Benchmark Issue

reactor-netty has issues with http 1.0 and hence ab. reactor/reactor-netty#21

The underlying problem is a bug in Reactor Netty: it does not support HTTP 1.0, so ab results are inaccurate for Spring Cloud Gateway.

Correct Benchmark Approach

Spring officials recommend using wrk for performance testing and provide a dedicated benchmark project called spring-cloud-gateway-bench , which compares three gateways:

Spring Cloud Gateway

Zuul (1.x)

Linkerd

The test setup uses a simple static server written in Go; each gateway proxies the same endpoint, and throughput is measured in requests per second (RPS).

Results

Spring Cloud Gateway – 32,213 RPS

Zuul – 20,800 RPS

Linkerd – 28,051 RPS

Conclusion: Spring Cloud Gateway delivers 1.6× the throughput of Zuul and is faster than Linkerd.

Outlook

The Zuul referenced is the blocking‑IO Zuul 1.x version.

Spring Cloud Gateway is a promising project with easy onboarding and strong features.

Linkerd, implemented in Scala, is a production‑grade service mesh.

Zuul 2.x, based on Netty and supporting long connections, is not yet integrated into Spring Cloud.