Chapter 3 - Memory Management, Recent Systems

1
Chapter 3 - Memory Management, Recent Systems
CIS106 Microcomputer Operating Systems Gina
Rue CIS Faculty
Ivy Tech State College Northwest Region 01
2
Memory Manager

• Early memory allocation schemes
• required storing entire programs in main memory
  in contiguous locations
• Causing problems -
• fragmentation
• overhead of relocation

3
Memory Manager

• More sophisticated memory allocation schemes
• remove the restriction of storing programs
  contiguously
• eliminate the need for an entire program reside
  in memory during execution

4
Memory Manager

• Four types of more sophisticated memory
  allocation schemes
• paged
• demand paging
• segmented
• segmented/demand paged

5
Paged Memory Allocation

• Based on concept of dividing each incoming job
  into pages of equal size
• page size can be the same as memory block size,
  same as the sections of a disk (sector) where a
  job is stored
• page frames are sections of main memory

6
Paged Memory Allocation

• One sector will hold one page of job instructions
  and fit into one page frame of memory
• Before executing a program, the Memory Manager
  prepares it by
• determining the of pages in the program
• locating enough empty page frames in main memory
• loading all of the programs into them (in
  static paging the pages need not be contiguous)

7
Paged Memory Allocation

• Memory Manager uses tables to keep track of how a
  jobs pages fit into page frames in memory
• Job Table (JT)
• Page Map Table (PMT)
• Memory Map Table (MMT)
• All three tables reside in the part of main
  memory thats reserved for the OS

8
Paged Memory Allocation

• Job Table (JT)
• Contains two entries for each active job
• Job size
• memory location where the Page Map Table (PMT) is
  stored
• This dynamic list grows as jobs are loaded into
  the system and shrinks as theyre later completed

9
Paged Memory Allocation

• Page Map Table (PMT)
• Each active job has its own PMT that contains
• page its corresponding page frame memory
  address
• one entry per page
• page s are sequential (Page 0, Page 1, Page 3)
• The first entry is the PMT lists the page frame
  memory address for Page 0

10
Paged Memory Allocation

• Memory Map Table (MMT)
• One entry for each page frame
• lists the location and free/busy status for each
  one
• at compilation time every job is divided into
  pages
• Example
• Job1 - Page 0 contains 1st hundred lines
• Displacement, or offset, of lines (how far away a
  line is from the beginning of its page) is a
  factor used to locate that line within its page
  frame

11
Paged Memory Allocation

• Each time and instruction is executed or a data
  value is used, the OS or (hardware) must
• translate the job space address
• which is relative
• into a physical address
• which is absolute, also known as resolving the
  address or address resolution

12
Paged Memory Allocation

• Advantages
• allows jobs to be allocated in noncontiguous
  memory locations
• memory is used more efficiently, more jobs can
  fit into main memory
• Disadvantages
• overhead is increased
• internal fragmentation
• difficult to determine best page size for optimal
  use of resources

13
Demand Paging

• First widely used scheme to load only part of the
  program into memory for processing
• Jobs are divided into equal sized pages that
  initially reside in secondary storage
• As jobs begin to run, its pages are brought into
  memory only as they are needed
• While one section of a job is being processed,
  all other modules are idle

14
Demand Paging

• Has made virtual memory widely available
• Allows users to run jobs with less memory
• Gives the appearance of an almost-infinite/non-inf
  inite memory
• Successful implementation requires use of high
  speed direct access storage device that can work
  directly with the CPU

15
Demand Paging

• How and when are pages passed (or swapped )?
• depends on predefined policies that determine
  when to make room for needed pages and how to do
  so
• OS relies on tables (JT,PMT,MMT) to implement the
  algorithm

16
Demand Paging

• To move a new page, a resident page must be
  swapped back into secondary storage requires
  close interaction between
• hardware components
• software algorithms
• policy scheme
• When a page is in secondary storage, but not in
  memory, the OS takes over to solve the problem
• Page Interrupt Handler if all page frames are
  busy, decides which page will be swapped out

17
Demand Paging

• Solution to inefficient memory utilization, but
  not problem free
• Thrashing occurs
• when excessive amount of page swapping between
  main memory secondary storage
• when page removed from memory but is called back
  shortly after
• Page Fault
• a failure to find a page in memory

18
Page Replacement Policies Concepts

• Crucial to the efficiency of the system the
  algorithm must be carefully selected
• Two most well known algorithms
• FIFO first-in first-out policy
• based on theory best page to remove is the one
  that has been in memory longest
• LRU last-recently used policy
• chooses the last recently accessed to be swapped
  out

19
Page Replacement Policies Concepts

• Two most well known algorithms
• FIFO first-first out policy
• There is no guarantee that buying more memory
  will always result in better performance (FIFO or
  Beladys anomaly)
• LRU last-recently used policy
• Stack algorithm removal policy which functions in
  a way that increasing main memory will cause
  either a decrease in or the same of pages
  interrupts

20
Page Replacement Policies Concepts

• Mechanics of Paging
• To determine which pages will be swapped, Memory
  Manager needs specific information included in
  the PMT

21
Page Replacement Policies Concepts

• PMT includes 3 bits
• status-indicates whether the page is currently in
  memory
• reference-indicates whether the page has been
  called (referenced) recently
• modified-indicates whether the contents of the
  page have been altered is used to determine if
  the page must be written to secondary storage

22
Page Replacement Policies Concepts

• Swapping pages
• FIFO algorithm uses only the modified bit and
  status bit
• LRU looks at all three (status, modified,
  referenced bit) before deciding which pages to
  swap out

23
Page Replacement Policies Concepts

• The Working Set
• Improved performance of page demanding schemes
• Is the set of pages residing in memory that can
  be accessed directly without incurring a page
  fault

24
Segmented Memory Allocation

• Segmenting is based on programmers use of
  structuring their programs in modules (logical
  groupings of code)
• each job is divided into several segments
• subroutine is an example of a logical group
• size of memory segments are each different
• when a program is compiled or assembled, the
  segments are set up according to the programs
  structural modules
• each segment is numbered and a Segment Map Table
  (SMT) is generated for each job

25
Segmented Memory Allocation

• The Memory Manager needs to keep track of
  segments in memory with three tables combining
  aspects of dynamic demand paging memory
  management
• The Job Table lists every job in process (one for
  the whole system
• The Segment Map Table lists details about each
  segment (one for each job)
• The Memory Map Table monitors the allocation of
  main memory (one for the whole system)

26
Segmented/Demand Paged Memory Allocation

• A combination of segmentation and demand paging
• offers the logical benefits of segmentation
• has physical benefits of paging
• This scheme requires 4 tables
• JT lists every job in process
• SMT lists details about each segment
• PMT lists details about every page
• MMT monitors the allocation of the page frames in
  main memory
• Associative memory is used to speed up the process

27
Virtual Memory

• Gives the user the appearance programs are being
  completely loaded in main memory during entire
  processing time
• During 2nd generation, programmers started
  dividing their programs into sections that
  resembled working sets, segments called
  overlays
• the programs could begin with only the first
  overlay loaded into memory

28
Virtual Memory

• Works well in multiprogramming environment
• Waiting time isnt lost
• CPU moves on to another job
• Increased programming techniques for development
  of large software systems because individual
  pieces can be created independently and linked
  together later

29
Virtual Memory

• Advantages
• job size no longer restricted
• memory used more efficiently
• unlimited amount of multiprogramming
• eliminates external fragmentation minimizes
  internal fragmentation
• allows the sharing of code and data
• facilitates dynamic linking of program segments

30
Virtual Memory

• Disadvantages
• Increased hardware costs
• Increased overhead for handling paging interrupts
• Increased software complexity to prevent
  thrashing

31
Summary

• Memory Manager has the task of allocating memory
  to each job to be executed and reclaiming it when
  execution is completed
• Each scheme was designed to address a different
  set of processing problems, but when some were
  solved, others were created