NGINX release 1.9.1 introduces a new feature that enables use of the
SO_REUSEPORT socket option, which is available in newer versions of many operating systems, including DragonFly BSD and Linux (kernel version 3.9 and later). This socket option allows multiple sockets to listen on the same IP address and port combination. The kernel then load balances incoming connections across the sockets.
Editor’s note – For NGINX Plus users, this feature is supported in NGINX Plus Release 7 (R7) and later. For an overview of all the new features in that release, see Announcing NGINX Plus R7 on our blog.
SO_REUSEPORT socket option has many potential real-world applications. Other services can use it for easy rolling upgrades of executables (NGINX already supports rolling upgrades through different means). For NGINX, enabling this socket option can improve performance in certain scenarios by reducing lock contention.
As depicted in the figure, when the
SO_REUSEPORT option is not enabled, a single listening socket notifies workers about incoming connections, and each worker tries to take a connection.
SO_REUSEPORT option enabled, there are multiple socket listeners for each IP address and port combination, one for each worker process. The kernel determines which available socket listener (and by implication, which worker) gets the connection. This can reduce lock contention between workers accepting new connections, and improve performance on multicore systems. However, it can also mean that when a worker is stalled by a blocking operation, the block affects not only connections that the worker has already accepted, but also connection requests that the kernel has assigned to the worker since it became blocked.
Configuring Socket Sharding
To enable the
SO_REUSEPORT socket option, include the new
reuseport parameter to the
listen directive for HTTP or TCP (stream module) traffic, as in these examples:
listen 80 reuseport;
listen 12345 reuseport;
reuseport parameter also disables the
accept_mutex directive for the socket, because the mutex is redundant with
reuseport. It can still be worth setting
accept_mutex if there are ports on which you don’t set
Benchmarking Performance with
I ran a wrk benchmark with 4 NGINX workers on a 36-core AWS instance. To eliminate network effects, I ran both client and NGINX on localhost, and also had NGINX return the string
OK instead of a file. I compared three NGINX configurations: the default (equivalent to
accept_mutex on), with
accept_mutex off, and with
reuseport. As shown in the figure,
reuseport increases requests per second by 2 to 3 times, and reduces both latency and the standard deviation for latency.
I also ran a related benchmark with the client and NGINX on separate hosts and with NGINX returning an HTML file. As shown in the following table, with
reuseport the decrease in latency was similar to the previous benchmark, and the standard deviation decreased even more dramatically (almost ten-fold). Other results (not shown in the table) were also encouraging. With
reuseport, the load was spread evenly across the worker processes. In the default condition (equivalent to
accept_mutex on), some workers got a higher percentage of the load, and with
accept_mutex off all workers experienced high load.
|Latency (ms)||Latency stdev (ms)||CPU Load|
In these benchmarks, the rate of connection requests is high but the requests don’t require extensive processing. Other preliminary testing also indicates that
reuseport improves performance the most when traffic matches this profile. (The
reuseport parameter is not available on the
listen directive in the
reuseport to determine whether it improves performance in your NGINX deployment, rather than applying it wholesale. For some tips on testing NGINX performance, check out Konstantin Pavlov’s talk at nginx.conf 2014.
Thanks to Yingqi Lu at Intel and Sepherosa Ziehau, who each contributed a solution to the NGINX project that enables use of the
SO_REUSEPORT socket option. The NGINX team combined ideas from both contributions to create what we believe is an ideal solution.