Redis [QUESTION] redis-benchmark shows different result between running as process and running in kubernetes pod (with hostNetwork)

Pod: hostNetwork, no cpu/mem limits

testing command: redis-benchmark -n 100000 -t set,get -h $IP -p 6379 -c 1

process gain 20k, pod gains 15

Any idea to help debug this?

Comment From: itamarhaber

Hello @YanzhaoLi

Have you looked at https://redis.io/topics/latency?

That said, I don't think the drop is related to Redis but rather to the containerization, but I'm not a k8s expert.

Comment From: filipecosta90

Hi there @YanzhaoLi as @itamarhaber stated this is very likely an infra-induced related issue and not a redis-server/redis-benchmark issue IMHO. Furthermore, notice that is extremely unlikely that we can pinpoint something rather than the underlying infrastructure (network) to be the cause of those latency differences ( CPU for sure is not the bottleneck given we're testing a single client benchmark ). Have you tried using netperf/iperf to measure the underlying network latency/BW?