Nuke can render images locally on your workstation (see Output (Write) Nodes) or it can be setup to render images on a network render farm (see Render Farms). Before rendering, make sure that you are using the appropriate file name syntax (see File Name Conventions for Rendered Images) and verify that your project settings have the correct output format and proxy format selected (see Render Resolution and Format).
By default, if you choose to render frame 15, the resulting file is numbered accordingly, for example image.0015.rgb. However, you can change this behavior via expressions, specified start frames, and constant offsets (see Changing the Numbering of Rendered Frames).
Sometimes, you may want to render an image just to read the rendered image back in (see Using a Write Node to Read in the Rendered Image). Because reading the output from a file is faster than calculating its output by processing the node tree upstream, this can speed up large projects.
NOTE: Scripts that require Nuke to load a large number of files concurrently (for example, by having hundreds of Read nodes followed by TimeBlurs) may exceed the number of available file handles per process, causing problems when rendering scripts.
Nuke itself supports up to 2048 file handles on all systems; however, you may need to increase the file handle limit on your system.
On Mac OS X, you can increase the default limit of 256 (depending on your version) by entering the following command from the Terminal and then running Nuke from the same Terminal session:
ulimit -Sn 2048