How many requests per second per web server?

Question

PHP

It is interesting to know how many requests per second web servers issue during rush hour?
It is clear that all this is conditional, it depends on a huge number of different things: hardware, server settings, scripts, databases.
I’ll write about myself
nginx + php-fpm
Intel Quad Core Xeon X3440
16 GB of memory
php 5.5
also the server runs Redis
database separately, statics separately
the server works only with php
150 requests per second
after scaling to 2 servers increased to 220, but already on 2x

Answer 1

I made 2600 requests in 6.6 seconds on my page, which pulls data from the database, all requests were processed, the database managed to record the IP of all "visitors"

Answer 2

What do you think requests are?

Answer 3

Hm. Probably, at the peak it was up to 8k RPS per server. And so, in normal mode, 500-800 rps flew there per car.
Actually, more than 5 dozen machines in different DCs (something around 2xE5645, 64GiB ram).

Answer 4

In fact, RPS is highly dependent on the application. How well it is written. From the server settings as well (for example, how many people think / know about the possibility of using a request queue between nginx and php-fpm).
For an online store on a machine with 7Gb (the store took something around 2Gb), the RAM in the php-fpm + postgresql + redis bundle received 300 RPS long-term, i.e. constant stable speed without 500s, and 600 at the peak, but some requests still failed.

Answer 5

Peak 500rps. As a mongo database, on the same host, from 2 to 4 reads are performed, 1 record per request.
Server 2xE5-2650, 64GB. CPU utilization at the peak of 10%, 130 workers.