Outline for 2/9/98

• Last time: Replacement algorithms - to Clock
• Questions?
• Administrative:
• Midterm coming Friday
• Objective today: Replacement algorithms that allow growing and shrinking frame allocations, Thrashing

• Not requiring each process to live within a fixed number of frames, replacing only its own pages.
• Can apply previously mentioned algorithms globally to victimize any process's pages
• Algorithms that make number of frames explicit.

• Page faulting is dominating performance
• Causes:
• Memory is overcommited - not enough to hold locality sets of all processes at this level of multiprogramming
• Lousy locality of their programs
• Positive feedback loops reacting to paging I/O rates
• Load control is important (how many slices in frame pie?) LT/RT strategy (limit # processes in load phase)

• Arguments for larger pages
• Page table size (fewer PTEs for large pages)
• Disk performance*
• Number of faults to load working set.
• Arguments for smaller pages
• Minimize internal fragmentation (ave 1/2 of last page is wasted)
• Matching spatial locality of programs (procedures tend to be much smaller than page)

Page Size Example

• Let there be physical memory for 1/2 this address space.
• Smaller page size
  AWCYAWCY...
• Larger page size
  AWBXCYDZ ...

Disk Characteristics*

Access time =
Seek time +
Rotational latency +
Transfer time

Larger pages amortize the overhead of getting to the data over more data transferred (good assuming its useful data to have moved in)