select vs. poll vs. epoll

All 3 are used for I/O multiplexing: monitor multiple file descriptors to see whether I/O is possible on any of them.

epoll was meant to replace the older POSIX select and poll system calls.

Complexity and Scalability

• the older system calls operate in O(n) time: every time you call select or poll, the kernel needs to check from scratch whether your file descriptors are available for writing. The kernel doesn’t remember the list of file descriptors it’s supposed to be monitoring.
• epoll operates in O(1) time. It doesn't do the linear scan over all the watched descriptors. epoll uses a red-black tree (RB-tree) data structure to keep track of all file descriptors that are currently being monitored. epoll "scales well to large numbers of watched file descriptors."

Availablity and Portability

• select and poll are available on any Unix system.
• epoll is Linux specific (available after version 2.5.44).
• poll is a POSIX standard interface, so use that when portability is required.

poll vs select

Given a list of file descriptors, they tell you which ones have data available to read/write to. select and poll fundamentally use the same code. poll returns a larget set of possible results for file descriptors like POLLRDNORM, POLLRDBAND, POLLIN, POLLHUP, POLLERR, while select just tells you there's input/output/error.

poll can perform better than select if you have a sparse set of file descriptors. poll takes a pollfd argument to specify which file descriptors to monitor; select takes bitsets and loops the whole range.

level-triggered vs edge-triggered

• level-triggered: get a list of every file descriptor you’re interested in that is readable.
• edge-triggered: get notifications every time a file descriptor becomes readable.

poll is only level-triggered, but epoll can be used as either edge- or level-triggered interface.

epoll

• epoll_create: start epolling
• epoll_ctl: tell the kernel file descriptors you’re interested in updates about.
• epoll_wait: wait for updates about the list of files you are interested in. epoll_wait() itself is a blocking interface. Whenever you call epoll_wait(), it blocks your thread/process until any of the monitored events happens on the registered descriptors. Either you ask epoll to monitor a non-blocking or blocking FD, epoll interface, itself, is still blocking-based

Where epoll is used

• node.js uses libuv (which was written for the node.js project)
• the gevent networking library in Python uses libev/libevent
• golang uses some custom code. This looks like it might be the implementation of network polling with epoll in the golang runtime – it’s only about 100 lines which is interesting. You can see the general netpoll interface,it’s implemented on BSDs with kqueue instead.
• Webservers also implement epoll: Every time the webserver accept a connection with the accept system call, and it gets a new file descriptor representing that connection. There may be thousands of connections open at the same time. You need to know when people send you new data on those connections, so you can process and respond to them.