# Benchmark results

We compare FizzGate with major gateway products on the market, using the same environment and conditions, and the test objects are all single nodes. The Mock interface simulates a delay of 20ms, and the message size is about 2K.

Categories Products 600 concurrency
QPS
600 concurrency
90% Latency(ms)
1000 concurrency
QPS
1000 concurrency
90% Latency(ms)
Backend services Direct access to backend services 23540 32.19 27325 52.09
traffic gateway kong
v2.4.1
15662 50.87 17152 84.3
Application Gateway fizz-gateway-community
v2.0.0
12206 65.76 12766 100.34
Application Gateway spring-cloud-gateway
v2.2.9
11323 68.57 10472 127.59
Application Gateway shenyu
v2.3.0
9284 92.98 9939 148.61

# Benchmark details

# Hardware environment

The server where the backend Mock service is located:

  • 4-core 8G memory Intel(R) Xeon(R) CPU E5-2650 v3 @ 2.30GHz * 4
  • Operating system CentOS Linux release 7.6.1810 (Core)
  • Kernel version Linux version 3.10.0-957.21.3.el7.x86_64

The server where the gateway node is located:

  • 4-core 8G memory Intel(R) Xeon(R) CPU E5-2650 v3 @ 2.30GHz * 4
  • Operating system CentOS Linux release 7.6.1810 (Core)
  • Kernel version Linux version 3.10.0-957.21.3.el7.x86_64

The stress testing tool occupies very few resources and is installed on the gateway node server.

# Stress testing tools

Stress measurement software: wrk

Concurrent connections: 600, 1000

# Test case description

  • Use the Mock service to simulate an interface that takes 20ms and responds to about 2k messages.
  • The log level of all tested gateways is WARN
  • All tested gateways do not enable additional plug-ins (such as authentication, flow control, etc.)
  • JVM maximum heap memory 4G
  • Each pressure test lasts 5 minutes

# Screenshot of stress test results

  • Direct access to backend services:

-Kong:

  • fizz-gateway-community:

  • spring-cloud-gateway:

  • shenyu: