Skip to content

One process starving the others

A backup job issues thousands of read requests. Interactive processes on the same machine become unresponsive when they touch the disk.

Which scheduling approach helps most?