Trying to understand Redis ping latency test vs Ping command Latency Test - redis

Question

I am trying to understand latency vs maximum number or requests that can be served per second.
What I understood RTT is time taken for message to reach destination and acknowledgement back to source. So I assume server can only serve maximum requests per second should not exceed more then sum of avg round trip in a give second. My local ping test shows as
> ping 127.0.0.1
rtt min/avg/max/mdev = 0.089/0.098/0.120/0.012 ms
on average it takes 0.098 ms just for network round trip, which means 10 ping req/ms. So I assume that in sequential order a client can only execute maximum of 10_000 req/sec. while it turns out I am wrong. redis-benchmark tool shows something different.
> redis-benchmark -t set -c 1 -h 127.0.0.1
====== SET ======
100000 requests completed in 2.53 seconds
1 parallel clients
3 bytes payload
keep alive: 1
100.00% <= 1 milliseconds
39588.28 requests per second
a single client is able to execute 39 req/ms while i am expecting maximum of 10req/ms.
Can anyone help me where I went wrong or misunderstood ?

Answer 1

Commands can be pipelined even when using a single logical client thread, meaning: you can send lots of requests before the first response comes back. Responses always come back in request order (unless you're using pub/sub), so a pipelining client simply needs to keep a queue of sent messages that have not yet seen responses, and pair responses to requests as they arrive.
So: you aren't strictly bound by latency, although that remains a useful number. The raw throughout number (bound by bandwidth and server capacity) is also meaningful, since it is often the case that you want to issue multiple commands.