The second question is particularly important in networked or time-sharing systems,
which regularly swap jobs (or pages of jobs) into memory and back to secondary storage
to accommodate the needs of many users. The problem is this: every time a job is
reloaded back into memory (or has pages swapped), it has to generate several page
faults until its working set is back in memory and processing can continue. It is a timeconsuming
task for the CPU, which can’t be processing jobs during the time it takes to
process each page fault, as shown in Figure 3.12.