Page Replacement

Note that the copying of the victim page from the frame back to disk doubles the extra time we spent during page faults (in addition to the time it takes to copy the page we want to bring from disk to memory.)

• As such, most algorithms will try evicting a page that hasn't been modified since the time it was brought from disk. This will spare the operating system from the need to copy the page to the disk (since it didn't change at all!)
• To know which pages changed or didn't change, the OS can store a bit per each page called modify bit or dirty bit. If the bit is 1, this means that a change was made to some data on this page. Otherwise, the bit will be 0.

A goal of the algorithms we'll see is not only to make page replacement less time consuming but to reduce the number of occurring page faults. Working with more frames should reduce page faults, too, as the next slide diagram shows.

After we discuss each algorithm, we'll demonstrate how it works by running it on a sample sequence of pages that a process might request to load. We call this sequence the reference string.