Why is paging out policy based on exploiting the concept of locality and working set likely to perform well?

Why is paging out policy based on exploiting the concept of locality and working set开发者_如何学JAVA likely to perform well?

---

Is this a homework question? It sounds like you don't really understand the question.

The "working set" is the set of memory pages used by a process. Obviously, you want to keep those pages in memory. However, without a Crystal Ball, it is hard to know what pages will be needed in the future. In general, the pages that have not been used recently are candidates to be swapped out.

Most PCs have "Uniform Memory Architecture" (UMA). That means that all memory is equally accessible. (except for special memory such as cache).

Some large, multi-processor systems have "Non-uniform Memory Architecture" (NUMA). On a NUMA system, memory may be associated with a particular processor or group of processors. For a processor to access its "local" memory is much faster than accessing "non-local" memory.

On NUMA systems, the paging policy needs to consider where physical memory is located before allocating it to a particular process. This is "locality".

On a UMA system, locality is probably not important.

Does that give you a foundation to think about the question?