Virtual Memory

1
Virtual Memory

• CS 147
• October 30, 2007
• Chris Stewart

2
What is Virtual Memory?

• Computer technique that gives application the
  impression it is working with contiguous memory
  when in fact the memory is physically fragmented
• Memory may also spill into physical storage (hard
  drives)?
• Not only limited to using hard drive space as
  extended RAM
• Memory extension is a consequence of virtual
  memory techniques

3
What is Virtual Memory?

• Virtual memory is, by definition, tricking
  applications into thinking the memory they are
  working with is contiguous
• Makes application programming easier for users,
  as management of virtual addresses is done at a
  low level

4
A pretty picture

• Shamelessly stolen from Wikipedia

5
History

• 1940s-1950s No virtual memory. Large programs
  had to contain logic for managing two-level
  storage and were responsible for moving data
  between secondary and primary storage.
• 1960s Virtual memory developed, debate over
  efficiency lasted until 1969
• Virtual memory became widespread from there
• Initially, memory was segmented and these
  segments could vary in size
• Was inefficient with larger segments

6
How does it work?

• Multiple methods of implementing virtual memory
• Almost all of them require a Page Table
• A page is a block of contiguous virtual memory
  addresses
• Virtual addresses are unique to the accessing
  program or process and map uniquely to a physical
  address
• Provides memory protection, among other benefits
• x86 Implementation Two modes Real and Virtual

7
How does it work?

• Real
• Address registers contain an integer that
  addresses some word or byte of RAM
• Memory is addressed sequentially
• Virtual
• Memory is divided into pages that are 4096 bytes
  long
• Pages may reside in any available RAM location
• Address Registers are divided into an index and
  an offset
• Addresses are resolved by looking up the index in
  the page table, which yields a physical address,
  and adding the offset

8
How does it work?

• Translation between Virtual and Physical
  Addresses is done in hardware.
• This component looks up the physical address
  corresponding to a particular virtual address and
  passes it to the CPU.
• Should there be no record of the virtual address
  being looked up, a page fault interrupt is
  generated. This interrupt is handled by a
  special part of the operating system.

9
How does it work?

• The paging supervisor is a part of the operating
  system responsible for creating and maintaining
  page files.
• If the translation hardware raises a page fault,
  the following process occurs
• Paging supervisor searches page file(s) on disk
• Page containing the required virtual address is
  read into memory
• Page table updated to reflect movement of page
• Translation hardware told to restart search
• Uses LRU ordering when determining which area of
  physical memory to overwrite

10
Paging

• Confusing terminology time!
• Paging is the process of saving inactive memory
  pages to disk and restoring them to physical
  memory when they are needed again.
• Pages do not necessarily return to the same
  physical memory location they left
• Applications are only aware of their virtual
  addresses
• Page table must be updated each time memory
  changes physical locations

11
Another Pretty Picture

• Which has also been shamelessly stolen from
  Wikipedia

12
So why do it?

• Extend memory onto hard disk, allowing programs
  access to more memory than is currently installed
  on the system
• Simply called page file on NT systems
• Called swap space on Unix and Unix-like systems
• Memory is protected applications are only aware
  of a small subset of virtual addresses, which
  makes memory conflicts less likely

13
Sources

• Wikipedia!