Virtual Memory

• Virtual memory is a memory management technique that gives applications the appearance of a large, continuous memory block, even with limited physical memory (RAM).
• It enables larger applications to run on systems with less RAM.
• The primary objective is to support multiprogramming.
• A running process does not need to be entirely in memory.
• Programs can be larger than the available physical memory.
• Virtual memory abstracts main memory, eliminating concerns about storage limitations.
• A memory hierarchy uses both RAM and disk, allowing processes to operate with only portions of their address space in RAM.
• Virtual memory provides an illusion of memory larger than the real memory, managed by a virtual memory manager.
• It is based on the noncontiguous memory allocation model, where the virtual memory manager removes components from memory to make room for others.
• The size of virtual storage is limited by the addressing scheme and secondary memory, not by main storage locations.

History of Virtual Memory

• Early computers used RAM and secondary memory for data storage, with magnetic core and drums respectively.
• In the 1940s and 1950s, computer memory was expensive and limited.
• Early programmers used overlaying when program was larger than available space, where code portions not currently used were moved.
• This led to the development of virtual memory.
• In 1956, German physicist Fritz-Rudolf Güntsch developed virtual memory.
• The first virtual memory system was created at the University of Manchester for the Atlas computer, using paging to map virtual addresses to main memory.
• The Atlas computer was built in 1959 and started working in 1962.
• The first computer with virtual memory was released by Burroughs Corp in 1961, using segmentation instead of paging.

How Virtual Memory Works

• Implemented using both hardware and software, it maps virtual addresses to physical addresses.
• Memory references within a process are logical addresses dynamically translated into physical addresses at run time.
• A process can be swapped in and out of main memory, occupying different places at different times.
• A process can be broken into pieces that don't need to be contiguous in main memory.
• This is achieved through dynamic run-time address translation and page/segment tables.
• Not all pages or segments need to be in main memory during execution; required pages are loaded when needed.
• Virtual memory is implemented using demand paging or demand segmentation.
• The Memory Management Unit (MMU), often in the CPU, manages virtual memory.
• The CPU generates virtual addresses that the MMU translates into physical addresses.

Types of Virtual Memory

• Paging divides memory into fixed-size blocks called pages.
• Segmentation divides memory into variable-size blocks called segments.

Paging

• When RAM is full, inactive pages are moved to a swap file on the hard drive acting as extended RAM.
• Page swapping is when pages are moved back into RAM when needed.
• This ensures the OS and applications have enough memory to run.

Demand Paging

• Loads pages into memory on demand (when a page fault occurs).
• If the CPU tries to access a page not in memory, it generates an interrupt.
• The OS puts the process in a blocking state and searches for the page in logical address space.
• The page is moved from logical address space to physical address space.
• Page replacement algorithms determine which page to replace in physical address space.
• The page table is updated.
• A signal is sent to the CPU to continue execution.

Page Fault Service Time

• The time taken to service a page fault.
• This includes all steps of finding, retrieving, and placing the page.
• m = Main memory access time.
• s = Page fault service time.
• p = Page fault rate.
• Effective memory access time = (p*s) + (1-p)*m

Page vs. Frame

• A page is a fixed-size block of data in virtual memory.
• A frame is a fixed-size block of physical memory in RAM where pages are loaded.

Segmentation

• Divides virtual memory into segments of different sizes.
• Segments not currently needed are moved to the hard drive.
• A segment table tracks each segment's status, including location and modification status.
• Segments are mapped into a process’s address space only when needed.

Combining Paging and Segmentation

• Memory is divided into pages, and segments are made up of multiple pages.
• The virtual address includes both a segment number and a page number.

Virtual Memory vs Physical Memory

What is Swapping?

• Swapping out means either removing all of a process's pages from memory or marking them for removal.
• Suspending a process ensures it is not runnable while swapped out.
• The system swaps the process back from secondary storage to main memory later.
• Thrashing occurs when a process spends too much time swapping pages in and out.

What is Thrashing?

• Occurs when the system spends more time swapping pages than executing instructions.
• The OS must manage memory well, because when the OS brings one page in, it must throw another out. If it throws out a page just before it is used, then it will just have to get that page again almost immediately.
• Results in poor system performance.
• Requires a good page replacement algorithm.
• CPU utilization is very high up to a point, but further increasing it causes it to drastically fall.

Causes of Thrashing

1. High Degree of Multiprogramming

   • Increasing processes reduce frames per process.
   • This leads to more frequent page faults and more CPU time wasted in swapping.

2. Lack of Frames

   • Insufficient frames for a process lead to frequent swapping.
   • A sufficient amount of frames must be allocated to each process in order to prevent thrashing.

Recovery of Thrashing

• Prevent thrashing by not bringing more processes into memory after a threshold.
• If thrashing occurs, suspend some processes using the mid-term scheduler.

Performance in Virtual Memory

• p = page fault rate (0 <= p <= 1)