What is Traefik?
Traefik is a modern seven-layer reverse proxy and load balancer, centered around a core concept — Service Discovery First.
The biggest difference from traditional reverse proxies (like Nginx) is that you don't need to manually write a large number of configuration files; Traefik proactively "observes" your runtime environment and automatically generates routing rules.
Because of the "service as configuration" feature, Traefik is especially suitable for these scenarios:
- Docker / Docker Compose
- Kubernetes（Ingress Controller）
- Microservices architecture
- Online services for small teams or independent developers
Simply put, if your service is "dynamic," Traefik is much less hassle-free than traditional reverse agents.
Benchmark
disposition
I want to thank exoscale.ch VincentBernat, who generously provided the infrastructure needed for benchmarking.
I tested it using 4 virtual machines, configured as follows:
32 GB of memory
8 CPU cores
10GB solid-state drive
Ubuntu 14.04 LTS 64-bit
Set up
A virtual machine is used to launch the benchmarking tool wrk
A virtual machine for Traefik (v1.0.0-beta.416) / nginx (v1.4.6)
Who is configuring two virtual machines for two backend servers in Go?
Each virtual machine has been tuned using the following restrictions:
Nginx
Below is the usage method for the Nginx configuration file /etc/nginx/nginx.conf:
Below are the Nginx virtual hosting configuration files used:
Traefik
The following are the files used for traefik.toml:
outcome
whoami：
wrk -t20 -c1000 -d60s -H "Host: test.traefik" --latency http://IP-whoami:80/bench
Running 1m test @ http://IP-whoami:80/bench
20 threads and 1000 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 70.28ms 134.72ms 1.91s 89.94%
Req/Sec 2.92k 742.42 8.78k 68.80%
Latency Distribution
50% 10.63ms
75% 75.64ms
90% 205.65ms
99% 668.28ms
3476705 requests in 1.00m, 384.61MB read
Socket errors: connect 0, read 0, write 0, timeout 103
Requests/sec: 57894.35
Transfer/sec: 6.40MB
nginx：
wrk -t20 -c1000 -d60s -H "Host: test.traefik" --latency http://IP-nginx:8001/bench
Running 1m test @ http://IP-nginx:8001/bench
20 threads and 1000 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 101.25ms 180.09ms 1.99s 89.34%
Req/Sec 1.69k 567.69 9.39k 72.62%
Latency Distribution
50% 15.46ms
75% 129.11ms
90% 302.44ms
99% 846.59ms
2018427 requests in 1.00m, 298.36MB read
Socket errors: connect 0, read 0, write 0, timeout 90
Requests/sec: 33591.67
Transfer/sec: 4.97MB
Traefik
wrk -t20 -c1000 -d60s -H "Host: test.traefik" --latency http://IP-traefik:8000/bench
Running 1m test @ http://IP-traefik:8000/bench
20 threads and 1000 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 91.72ms 150.43ms 2.00s 90.50%
Req/Sec 1.43k 266.37 2.97k 69.77%
Latency Distribution
50% 19.74ms
75% 121.98ms
90% 237.39ms
99% 687.49ms
1705073 requests in 1.00m, 188.63MB read
Socket errors: connect 0, read 0, write 0, timeout 7
Requests/sec: 28392.44
Transfer/sec: 3.14MB
conclusion
Traefik's speed is clearly slower than Nginx's, but the gap isn't huge: Traefik handles 28,392 requests per second, while Nginx handles 33,591 requests per second, with a performance ratio of 85%. For a young project, that's pretty good :)!
Some areas for improvement:
Use GO_REUSEPORT monitors
Each CPU core runs a separate server instance GOMAXPROCS=1 (benchmarks show Traefik has much more context switching than nginx).
Original:The hyperlink login is visible. |
Posted on11 minutes ago
|
|
|
