Chapter 8

[ 211 ]

Performance problems

When designing an application and configuring NGINX to deliver it, we expect it to

perform well. When we experience performance problems, however, we need to take a

look at what could cause them. It may be in the application itself. It may be our NGINX
configuration. We will investigate how to discover where the problem lies.

When proxying, NGINX does most of its work over the network. If there are any

limitations at the network level, NGINX cannot perform optimally. Network tuning

is again specific to the operating system and network that you are running NGINX
on, so these tuning parameters should be examined in your particular situation.

One of the most important values relating to network performance is the size of the

listen queue for new TCP connections. This number should be increased to enable

more clients. Exactly how to do this and what value to use depends on the operating
system and optimization goal.

• Linux
  vi /etc/sysctl.conf

net.core.somaxconn = 3240000

# sysctl -p /etc/sysctl.conf

• FreeBSD
  vi /etc/sysctl.conf

kern.ipc.somaxconn=4096

# /etc/rc.d/sysctl reload

• Solaris

  ndd -set /dev/tcp tcp_conn_req_max_q 1024

  
  

  ndd -set /dev/tcp tcp_conn_req_max_q0 4096