Troubleshooting Techniques

[ 212 ]

The next parameter to change is the size of the send and receive buffers. Note that

these values are for illustration purposes only— they may lead to excessive memory
usage, so be sure to test in your specific scenario.

• Linux
  vi /etc/sysctl.conf

net.ipv4.tcp_wmem = 8192 87380 1048576

net.ipv4.tcp_rmem = 8192 87380 1048576

# sysctl -p /etc/sysctl.conf

• FreeBSD
  vi /etc/sysctl.conf

net.inet.tcp.sendspace=1048576

net.inet.tcp.recvspace=1048576

# /etc/rc.d/sysctl reload

• Solaris

  ndd -set /dev/tcp tcp_xmit_hiwat 1048576

  
  

  ndd -set /dev/tcp tcp_recv_hiwat 1048576

  
  

You can also change these buffers in NGINX's configuration directly, so that they are
only valid for NGINX and not for any other software you are running on the machine.

This may be desirable when you have multiple services running, but want to ensure

that NGINX gets the most out of your network stack:

server {

listen 80 sndbuf=1m rcvbuf=1m;

}

Depending on your network setup, you will notice a marked change in performance.

You should examine your particular setup, though, and make one change at a time,
observing the results after each change. Performance tuning can be done on so many

different levels that this small treatment here does not do the subject justice. If you

are interested in learning more about performance tuning, there are a number of
books and online resources that you should take a look at.