Basic network performance testing
alpine:3.7
qperf from testing repo
kubernetes > 1.8.0
Change SERVER_ADDR in the client deployment config to your cluster's record.
Standard D2 v2 (2 vcpus, 7 GB memory)
iteration:1 t:16 latency = 171 us
iteration:2 t:39 latency = 225 us
iteration:3 t:37 latency = 205 us
iteration:4 t:41 latency = 174 us
iteration:5 t:37 latency = 210 us
iteration:6 t:11 latency = 184 us
iteration:7 t:48 latency = 237 us
iteration:8 t:13 latency = 178 us
iteration:9 t:36 latency = 198 us
iteration:10 t:10 latency = 187 us
iteration:11 t:16 latency = 171 us
iteration:12 t:32 latency = 167 us
iteration:13 t:34 latency = 162 us
iteration:14 t:33 latency = 221 us
iteration:15 t:55 latency = 197 us
iteration:16 t:55 latency = 213 us
iteration:17 t:31 latency = 202 us
iteration:18 t:49 latency = 226 us
iteration:19 t:35 latency = 176 us
iteration:20 t:14 latency = 218 us
kubectl apply -f ./service
antiafinity or daemonset make sure client is not on the same node as server.