High response time at high concurrent connections

Yichun Zhang (agentzh) agentzh at gmail.com
Mon Sep 30 02:25:13 UTC 2013


On Sun, Sep 29, 2013 at 3:50 PM, laltin wrote:
> I can handle ~1200 requests/seconds with 1000 concurrent connections and
> when I do the same benchmark with 100 concurrent connections I can again
> handle around 1200 requests/second but response time drops to ~80 ms.

What you're seeing is a very common phenomenon.

Tools like ab always try to load the target server to its extreme
throughput, i.e., the 1200 req/sec you're seeing is already the
throughput limit for that service in your server. Because you're
already at the throughput limit, you have to sacrifice response
latency for higher concurrency level otherwise you'll just have much
higher throughput which is impossible (remember you're already on the
limit? ;)).


More information about the nginx mailing list