In the previous example with dd(1), we saw that we can indeed impose a limit upon the maximum file size. An important question arises: what is the scope or granularity of the resource limit? Is it system-wide?

The short answer: no, it's not system-wide, it's process-wide, implying that the resource limits apply at the granularity of a process and not the system. To clarify this, consider two shells—nothing but the bash process—shell A and shell B. We modify the maximum file-size resource limit for shell A (with the usual ulimit -f <new-limit> command), but leave the resource limit for maximum file size for shell B untouched. If now they both use dd (as we did), we would find that the dd  process invoked within shell A would likely die with the 'File size limit exceeded (core dumped)' failure message, whereas the dd process invoked within shell B would likely continue and succeed (provided, of course, there's sufficient disk space available).

This simple experiment proves that the granularity of a resource limit is per process.