Test Setup

We create 10 mutating webhooks that intecepts configMap creation, then run batchNumber threads to create configMaps, each creating batchSize configMaps in serial. We measure the run time of the client.

Results

Latency in seconds for Direct connection between kube-apiserver and cluster:

#batches\batch size 1 10 100 1000
1 2.0 21.1
10 1.1 9.7
100 1.3 10.3

Latency in seconds for connection going through network proxy running in GRPC mode:

#batches\batch size 1 10 100 1000
1 24.9 243
10 12.0 89.0
100 16.8 101

Latency in seconds for connection going through network proxy running in HTTP-Connect mode:

#batches\batch size 1 10 100 1000
1 31.2 240
10 17.9 152
100 19.7 86

Latency for ssh connection: TODO

Use this test suite

TODO