Memory Management

Memory management
• We have seen how CPU can be shared by a set
of processes
– Improve system performance
– Process management
• Need to keep several process in memory
– Share memory
• Learn various techniques to manage memory
– Hardware dependent
 Memory management
What are we going to learn?
• Basic Memory Management: logical vs.
physical address space, protection, contiguous
memory allocation, paging, segmentation,
segmentation with paging.

• Virtual Memory: background, demand paging,

performance, page replacement, page
replacement algorithms (FCFS, LRU), allocation
of frames, thrashing.
 Background

• Program must be brought (from disk) into

memory
CPU
• Fetch-decode-execute cycle

• Memory unit only sees a stream of

addresses + read requests, or address + data
and write requests
• Sequence of memory addresses generated
by running program
 Logical vs. Physical Address Space

Logical address – generated by the CPU; also

referred to as virtual address CPU

Physical address – address seen by the memory

unit
• Logical address space is the set of all logical
addresses generated by a program
• Physical address space is the set of all
physical addresses generated by a program
 Background
Multiple processes resides in memory

• Protection of memory required to ensure

correct operation

1. Protect OS
2. Protect user processes
 Base and Limit Registers
• A pair of base and limit registers define
the logical address space
 Hardware Address Protection with Base and Limit Registers

• OS loads the base & limit reg.

• Privileged instruction
 Address Binding
• Process resides in main memory
• Associate each data element with memory address
• Further, addresses represented in different ways at
different stages of a program’s life
– Source code addresses usually symbolic
– Compiled code addresses bind to relocatable addresses
• i.e. “14 bytes from beginning of this module”
– Linker or loader will bind relocatable addresses to absolute
addresses
• i.e. 74014
Multistep Processing of a User
Program
 Binding of Instructions and Data to
Memory
• Address binding of instructions and data to memory
addresses can happen at three different stages
– Compile time: If memory location known a priori,
absolute code can be generated; must recompile code if
starting location changes
– Load time: Must generate relocatable code if memory
location is not known at compile time
– Execution time: If the process can be moved during its
execution from one memory segment to another
• Binding delayed until run time
• Need hardware support for address maps (e.g., base and limit
registers)
 Logical vs. Physical Address Space
Logical address – generated by the CPU; also referred to
as virtual address
CPU

Physical address – address seen by the memory unit

• Logical and physical addresses are the same in

compile-time and load-time address-binding
schemes;
• logical (virtual) and physical addresses differ in
execution-time address-binding scheme
• Logical address space is the set of all logical
addresses generated by a program
• Physical address space is the set of all physical
addresses generated by a program
 Memory-Management Unit (MMU)
• Hardware device that at run time maps virtual to physical address

• Many methods possible

• To start, consider simple scheme where the value in the

relocation register is added to every address generated by a user
process at the time it is sent to memory
– relocation register
– MS-DOS on Intel 80x86 used 4 relocation registers

• The user program deals with logical addresses (0 to max); it

never sees the real physical addresses (R to R+max)
– Say the logical address 25
– Execution-time binding occurs when reference is made to location
in memory
– Logical address bound to physical addresses
Dynamic relocation using a
relocation register

14000

Relocatable
code
 Hardware Support for Relocation
and Limit Registers

• Relocation registers used to protect user processes from each other, and from changing
operating-system code and data
• Relocation register contains value of smallest physical address
• Limit register contains range of logical addresses – each logical address must be less
than the limit register
• Context switch
• MMU maps logical address dynamically

• OS loads the reloc & limit reg.

• Privileged instruction
 Contiguous Allocation

Multiple processes resides in memory

 Contiguous Allocation

• Main memory usually divided into two

partitions:
– Resident operating system, usually held in high
memory
– User processes then held in low memory
– Each process contained in single contiguous
section of memory
 Contiguous Allocation (Cont.)
• Multiple-partition allocation
– Divide memory into several Fixed size partition
– Each partition stores one process
– Degree of multiprogramming limited by number of
partitions
– If a partition is free, load process from job queue
– MFT (IBM OS/360)
 Contiguous Allocation (Cont.)
• Multiple-partition allocation
– Variable partition scheme
– Hole – block of available memory; holes of various size are
scattered throughout memory
– Keeps a table of free memory
– When a process arrives, it is allocated memory from a hole large
enough to accommodate it
– Process exiting frees its partition, adjacent free partitions
combined
– Operating system maintains information about:
a) allocated partitions b) free partitions (hole)

OS OS OS OS OS

process 5 process 5 process 5 process 5

process 9 process 9
Hole process 8 process 10

process 2 process 2 process 2 process 2

 Dynamic Storage-Allocation Problem
How to satisfy a request of size n from a list of free holes?
Dynamic storage allocation problem
• First-fit: Allocate the first hole that is big enough

• Best-fit: Allocate the smallest hole that is big enough; must search
entire list, unless ordered by size
– Produces the smallest leftover hole

• Worst-fit: Allocate the largest hole; must also search entire list
– Produces the largest leftover hole
 Fragmentation
• Processes loaded and removed from memory
– Memory is broken into little pieces

• External Fragmentation – total memory space

exists to satisfy a request, but it is not contiguous

• First fit analysis reveals that given N blocks

allocated, 0.5 N blocks lost to fragmentation
– 1/3 may be unusable -> 50-percent rule
 Fragmentation (Cont.)
• Reduce external fragmentation by compaction
– Shuffle memory contents to place all free memory
together in one large block
– Compaction is possible only if relocation is dynamic,
and is done at execution time
• Change relocation reg.
– Cost

• Internal Fragmentation – allocated memory may

be slightly larger than requested memory; this size
difference is memory internal to a partition, but
not being used
 Paging
• Physical address space of a process can be noncontiguous;
– process allocates physical memory whenever the latter is available
• Divide physical memory into fixed-sized blocks called frames
– Size is power of 2, between 512 bytes and 16 Mbytes

• Divide logical memory into blocks of same size called pages

– To run a program of size N pages, need to find N free frames and load
program

• Backing store likewise split into pages

• Set up a page table to translate logical to physical addresses

• System keeps track of all free frames

 Paging Model of Logical and Physical Memory

page table to translate logical to physical

addresses
 Address Translation Scheme
• Address generated by CPU is divided into:
– Page number (p) – used as an index into a page table
• which contains base address of each frame in physical memory
– Page offset (d) – offset within a page
• combined with base address to define the physical memory address that
is sent to the memory unit

offset
page
page number page offset

p d
m-n n
– For given logical address space 2m and page size 2n
Paging Hardware
 Paging Example
Logical address 0
(0*4+0)
Logical address = 16 Physical address:
Page size=4 (5*4+0)=20
Physical memory=32
Logical address 3
(0*4+3)
Physical address:
(5*4+0)=23

Logical address 4
User’s view (1*4+0)
Physical address:
(6*4+0)=24
Run time address binding
Logical address 13
(3*4+1)
Physical address:
(2*4+1)
n=2 and m=4 32-byte
memory and 4-byte pages
 Paging
• External fragmentation??
• Calculating internal fragmentation
– Page size = 2,048 bytes
– Process size = 72,766 bytes
– 35 pages + 1,086 bytes
– Internal fragmentation of 2,048 - 1,086 = 962 bytes
• So small frame sizes desirable?
– But increases the page table size
– Poor disk I/O
– Page sizes growing over time
• Solaris supports two page sizes – 8 KB and 4 MB
• User’s view and physical memory now very different
– user view=> process contains in single contiguous memory space
• By implementation process can only access its own memory
– protection
 Page table entries
• Each page table entry 4 bytes (32 bits) long
• Each entry can point to 232 page frames
• If each frame is 4 KB
• The system can address 244 bytes (16TB) of
physical memory

Virtual address space 16MB.

Page table size?
• Process P1 arrives
• Requires n pages => n frames must be
available
• Allocate n frames to the process P1
• Create page table for P1
 Frame table Free Frames

Use’s view
System’s view

RAM RAM
Before allocation After allocation
Executable file and virtual address
Symbol table
Name address
SQR 0
a.out
SUM 4 Virtual address
space
 Implementation of Page Table
• For each process, Page table is kept in main memory
• Page-table base register (PTBR) points to the page table
• Page-table length register (PTLR) indicates size of the
page table
• In this scheme every data/instruction access requires two
memory accesses
– One for the page table and one for the data / instruction
• The two memory access problem can be solved by the
use of a special fast-lookup hardware cache called
associative memory or translation look-aside buffers
(TLBs)
Associative memory
 Associative Memory
• Associative memory – parallel search
Page # Frame #

• Address translation (p, d)

– If p is in associative register, get frame # out
– Otherwise get frame # from page table in memory
 Implementation of Page Table
• For each process, Page table is kept in main memory
• Page-table base register (PTBR) points to the page table
• Page-table length register (PTLR) indicates size of the page table
• In this scheme every data/instruction access requires two memory accesses
– One for the page table and one for the data / instruction
• The two memory access problem can be solved by the use of a special fast-lookup
hardware cache called associative memory or translation look-aside buffers (TLBs)

• TLBs typically small (64 to 1,024 entries)

• On a TLB miss, value is loaded into the TLB for faster access next time
– Replacement policies must be considered (LRU)
– Some entries can be wired down for permanent fast access

• Some TLBs store address-space identifiers (ASIDs) in each TLB entry – uniquely
identifies each process (PID) to provide address-space protection for that process
– Otherwise need to flush at every context switch
Paging Hardware With TLB
 Effective Access Time
• Associative Lookup =  time unit
– Can be < 10% of memory access time

• Hit ratio = 
– Hit ratio – percentage of times that a page number is found in the
associative registers; ratio related to size of TLB

• Consider  = 80%,  = 20ns for TLB search, 100ns for memory access

• Effective Access Time (EAT)

EAT = (100 + )  + (200 + )(1 – )

• Consider  = 80%,  = 20ns for TLB search, 100ns for memory access
– EAT = 0.80 x 120 + 0.20 x 220 = 140ns

• Consider better hit ratio ->  = 98%,  = 20ns for TLB search, 100ns for
memory access
– EAT = 0.98 x 120 + 0.02 x 220 = 122ns
 Memory Protection
• Memory protection implemented by associating protection bit
with each frame to indicate if read-only or read-write access is
allowed
– Can also add more bits to indicate page execute-only, and so on

• Valid-invalid bit attached to each entry in the page table:

– “valid” indicates that the associated page is in the process’ logical
address space, and is thus a legal page
– “invalid” indicates that the page is not in the process’ logical
address space
– Or use PTLR

• Any violations result in a trap to the kernel

 Valid (v) or Invalid (i)
Bit In A Page Table
14 bit address space (0 to 16383)
Page size 2KB
Process P1 uses only 0 to 10468
P2
Page 0 P1
Page 1
Page 2
Page 3

10240

Internal fragmentation Use of PTLR (length)

 Shared Pages Example
• System with 40 users
– Use common text editor
• Text editor contains 150KB code 50KB data (page size 50KB)
– 8000KB!
• Shared code
– One copy of read-only (reentrant) code shared among
processes (i.e., text editors, browsers)
• Code never changes during execution
• Only one copy of the editor in the memory
• Total memory consumption
– 40*50+150=2150KB
Shared Pages Example
 Data share: example
writer.c
reader .c
int main()
{ int main()
{
int shmid,f,key=3,i,pid;
char *ptr; int shmid,f,key=3,i,pid;
char *ptr;
shmid=shmget((key_t)key,100,IPC_CREAT|0666);
ptr=shmat(shmid,NULL,0); shmid=shmget((key_t)key,100,IPC_CREAT|0666);
printf("shmid=%d ptr=%u\n",shmid, ptr); ptr=shmat(shmid,NULL,0);
strcpy(ptr,"hello"); printf("shmid=%d ptr=%u\n",shmid, ptr);
i=shmdt((char*)ptr); printf("\nstr %s\n",ptr);
}
}

ptr
Shared
memory
 Structure of the Page Table
• Memory requirement for page table can get huge using straight-
forward methods
– Consider a 32-bit logical address space as on modern computers
– Page size of 4 KB (212)
– Page table would have 1 million entries 220 (232 / 212)
– If each entry is 4 bytes -> 4 MB of physical address space / memory for
page table alone
• That amount of memory used to cost a lot
• Don’t want to allocate that contiguously in main memory

• Hierarchical Paging

• Hashed Page Tables

• Inverted Page Tables

 Hierarchical Page Tables
• Break up the page table into multiple pages

• We then page the page table, and store it in

frame

• Create outer page table

• A simple technique is a two-level page table

Two-Level Page-Table Scheme
 Two-Level Paging Example
• A logical address (on 32-bit machine with 4KB page size) is
divided into:
– a page number consisting of 20 bits
– a page offset consisting of 12 bits
• Since the page table is paged, the page number is further
divided into:
– a 10-bit page number
– a 10-bit page offset
• Thus, a logical address is as follows:

page number page offset

p1 p2 d
10 10 12

• where p1 is an index into the outer page table, and p2 is the

displacement within the page of the inner page table
 Two-Level Page-Table Scheme
Each divided page table
size=210 *4bytes=4KB
=Page size

d
p1

p2

Pentium II
 Address-Translation Scheme

Pentium II
 64-bit Logical Address Space
• Even two-level paging scheme not sufficient
• If page size is 4 KB (212)
– Then page table has 252 entries
– If two level scheme, inner page tables could be 210 4-byte entries
– Address would look like
inner page
outer page page offset
p1 p2 d
42 10 12

– Outer page table has 242 entries or 244 bytes

– One solution is to add a 2nd outer page table
– But in the following example the 2nd outer page table is still 234
bytes in size
• And possibly 4 memory access to get to one physical memory location
 Three-level Paging Scheme

SPARC (32 bits), Motorola 68030 support three and four level paging respectively
 Hashed Page Tables
• Common in virtual address spaces > 32 bits

• The page number is hashed into a page table

– This page table contains a chain of elements hashing to the same
location

• Each element contains (1) the page number (2) the value of the
mapped page frame (3) a pointer to the next element

• Virtual page numbers are compared in this chain searching for a

match
– If a match is found, the corresponding physical frame is extracted
Hashed Page Table
 Inverted Page Table
• Rather than each process having a page table and
keeping track of all possible logical pages, track all frames

• One entry for each frame

• Entry consists the page number stored in that frame,

with information about the process that owns that page

• Decreases memory needed to store each page table,

– but increases time needed to search the table when a page
reference occurs
 Inverted Page Table Architecture
64 bit UltraSPARC, PowerPC,

Address space ID
 Swapping
A process, or a portion of a process, can be swapped temporarily out of memory to a
backing store and then brought back into memory for continued execution

Expensive

Advantages :
Allows physical memory to be oversubscribed, so that the system can accommodate more
processes than there is actual physical memory to store them.
Idle or mostly idle processes are good candidates for swapping; any memory that has been
allocated to these inactive processes can then be dedicated to active processes.
Medium Term Scheduling
disk Swap Space

Swapper
 Swapping with Paging

• When we want to
execute a process, swap
in

• Instead of swap in entire

process, load page

• Pager
Swapping with Paging
 Segmentation
• Memory-management scheme that supports user view of
memory
• A program is a collection of segments
– A segment is a logical unit such as:
main program Compiler generates the
procedure segments
Loader assign the seg#
function
method
object
local variables, global variables
common block
stack
symbol table
arrays
Executable file and virtual address
Symbol table
Name address
SQR 0
a.out
SUM 4 Virtual address
space
User’s View of a Program
User specifies each address
by two quantities
(a) Segment name
(b) Segment offset

Logical address contains the

tuple
<segment#, offset>

• Variable size segments without order

• Length=> purpose of the program
• Elements are identified by offset
 Logical View of Segmentation
Logical address <segment-number, offset>
1

4
1

3 2
4

Logical
address
space user space physical memory space

• Long term scheduler finds and allocates memory for all segments of a program
• Variable size partition scheme
Memory image
Executable file and virtual address
Symbol table
Name address
SQR 0
a.out
SUM 4 Virtual address
space

Paging view
0 Load 0
4 ADD 4

Segmentation view
<CODE, 0> Load <ST,0>
<CODE, 2> ADD <ST,4>
 Segmentation Architecture
• Logical address consists of a two tuple:
<segment-number, offset>
• Segment table – maps two-dimensional logical address
to physical address;
• Each table entry has:
– base – contains the starting physical address where the
segments reside in memory
– limit – specifies the length of the segment
• Segment-table base register (STBR) points to the
segment table’s location in memory
• Segment-table length register (STLR) indicates number
of segments used by a program;
segment number s is legal if s < STLR
Example of Segmentation
Segmentation Hardware
Example of Segmentation
 Segmentation Architecture

• Protection
• Protection bits associated with segments
– With each entry in segment table associate:
• validation bit = 0  illegal segment
• read/write/execute privileges
• Code sharing occurs at segment level
• Since segments vary in length, memory allocation is
a dynamic storage-allocation problem
– Long term scheduler
– First fit, best fit etc
• Fragmentation
 Segmentation with Paging
Key idea:
Segments are splitted into multiple pages

Each page is loaded into frames in the memory

 Segmentation with Paging
• Supports both segmentation and segmentation with paging
– Each segment can be 4 GB
– Up to 16 K segments per process S(13) G(1) P(2)
– <selector(16), offset (32)>
– Divided into two partitions
• First partition of up to 8 K segments are private to process (kept in local
descriptor table LDT)
• Second partition of up to 8K segments shared among all processes (kept in
global descriptor table GDT)

• CPU generates logical address (six Segment Reg.)

– Given to segmentation unit
• Which produces linear (logical) addresses
– Physical address 32 bits
– Linear address given to paging unit Intel 80386
• Which generates physical address in main memory
• Paging units form equivalent of MMU IBM OS/2
• Pages sizes can be 4 KB
 Logical to Physical Address
Translation in Pentium 000
000
Each segment works as a
separate logical address space Segment 2
Segment 1

max1 max2

Page table=220
entries
 Example: The Intel Pentium
Segment #3

8 bytes Segment register

Segment table

Logical address of a
segment

Page table of
Segment #3
Intel Pentium Segmentation
Pentium Paging Architecture
Pages maybe on
disk, valid/invalid
bit
 Problem 1
Consider a system with byte-addressable memory, 32 bit logical addresses, 4 kilobyte
page size and page table entries of 4 bytes each. Compute the size of the page table in
the system in megabytes

Consider a machine with 64 MB physical memory and a 32-bit virtual address space. If
the page size is 4 KB, what is the approximate size of the page table?
 Problem 2

In a virtual memory system, size of virtual address is 32-bit, size of physical

address is 30-bit, page size is 4 Kbyte and size of each page table entry is 32-bit.
The main memory is byte addressable. Compute the maximum number of bits
that can be used for storing protection and other information in each page table
entry.
 Problem 3
Consider a system with 2 level paging, where the page table is divided into 8k
pages of size 16kB. The memory is byte addresable
The physical address space is 128MB which is divided into 4KB frames.
The page table entry of outer page table is 32 bits and page table entry size of
inner page table is 64bits.
Show the logical and physical address format.
What will be the page table size of inner and outer page table?
 Problem 4
Consider the segment table of a process

Process generates the following virtual addresses. Translate to physical address

Virtual Memory
• Process P1 arrives
• Requires n pages => n frames must be
available
• Allocate n frames to the process P1
• Create page table for P1
 Background
• Code needs to be in memory to execute, but
entire program rarely used
– Error code, unusual routines, large data structures
• Entire program code not needed at same time
• Consider ability to execute partially-loaded
program
– Program no longer constrained by limits of physical
memory
– programs could be larger than physical memory
– More processes can be accommodated
 Background
• Example
– 40 frames in memory
– 8 processes each needs 10 pages
– 5 of them rarely been used
• Two options
– Run 4 processes (10 pages)
– Run 8 processes (5 pages)
• Increase the degree of multiprogramming
– Over allocating memory
 Classical paging
• Process P1 arrives
• Requires n pages => n frames must be
available
• Allocate n frames to the process P1
• Create page table for P1

Allocate < n frames

 Virtual memory
• Virtual memory – separation of user logical memory from
physical memory
– Extremely large logical space is available to programmer
– Concentrate on the problem
• Only part of the program needs to be in memory for
execution
– Logical address space can therefore be much larger than physical
address space
– Starts with address 0, allocates contiguous logical memory
– Physical memory
• Collection of frame

• Virtual memory can be implemented via:

– Demand paging
– Demand segmentation
 Virtual Memory That is
Larger Than Physical Memory

Large virtual
space

Small memory
 Demand Paging
• Bring a page into memory only when it is needed

• Lazy swapper – never swaps a page into memory

unless page will be needed
– Swapper that deals with pages is a pager

• Less I/O needed, no unnecessary I/O

– Less memory needed
– More users

• Page is needed  reference to it Valid address

information is available
– invalid reference  abort in PCB
– not-in-memory  bring to memory
 Page Table When Some Pages
Are Not in Main Memory

Pager loads few necessary pages in

memory
 Valid-Invalid Bit
• With each page table entry a valid–invalid bit is associated
(v  in-memory – memory resident, i  not-in-memory)
• Initially valid–invalid bit is set to i on all entries
• Example of a page table snapshot:
Frame # valid-invalid bit
v
v
v
v
i page table

….

ii Disk
address
• During address translation, if valid–invalid bit in page table entry
is i  page fault
 Page Fault
• If the page in not in memory, first reference to that page will trap to
operating system:
page fault

1. Operating system looks at PCB to decide:

– Invalid reference  abort
– Just not in memory (load the page)
2. Get empty frame
3. Swap page into frame via scheduled disk operation
4. Reset page table to indicate page now in memory
Set validation bit = v
5. Restart the instruction that caused the page fault
Steps in Handling a Page Fault
Check
PCB
 Steps in the ISR
• In Demand Paging
1. Trap to the operating system
2. Save the user registers and process state
3. Determine that the interrupt was a page fault
4. Check that the page reference was legal and determine the location of the page on the disk
5. Get a free frame
6. Issue a read from the disk to a free frame:
1. Wait in a queue for this device until the read request is serviced
2. Wait for the device seek and/or latency time
3. Begin the transfer of the page to a free frame
7. While waiting, allocate the CPU to some other user
8. Receive an interrupt from the disk I/O subsystem (I/O completed)
9. Save the registers and process state of the running process
10. Determine that the interrupt was from the disk
11. Correct the page table and other tables to show page is now in memory
12. Wait for the CPU to be allocated to this process again
13. Restore the user registers, process state, and new page table, and then resume the interrupted
instruction
 Pure Demand Paging
• Extreme case – start process with no pages in memory
– OS sets instruction pointer to first instruction of process, non-
memory-resident -> page fault
– Swap in that page
– Pure demand paging
• Actually, a given instruction could access multiple pages
(instruction + data) -> multiple page faults
– Pain decreased because of locality of reference
• Hardware support needed for demand paging
– Page table with valid / invalid bit
– Secondary memory (swap device with swap space)
– Instruction restart after page fault
 Page and Frame Related Algorithms

• Page-replacement algorithm
– Which frames to replace
– Want lowest page-fault rate

• Frame-allocation algorithm determines

– How many frames to give each process
 What Happens if There is no Free Frame?
• Example
– 40 frames in memory
– 8 processes each needs 10 pages
– 5 of them never used
• Two options
– Run 4 processes (10 pages)
– Run 8 processes (5 pages)
• Increase the degree of multiprogramming
– Over allocating memory

• Page fault
– No free frame
– Terminate? swap out? replace the page?

• Page replacement – find some page in memory, not really in use, page it out

– Performance – want an algorithm which will result in minimum number of page faults

• Same page may be brought into memory several times

 Need For Page Replacement
P1

P2
 Need For Page Replacement
P1

P2

PC
 Basic Page Replacement
1. Find the location of the desired page on disk

2. Find a free frame:

- If there is a free frame, use it
- If there is no free frame, use a page replacement algorithm to
select a victim frame (of that process)
- Write victim frame to disk

3. Bring the desired page into the (newly) free frame; update the page
and frame tables

4. Continue the process by restarting the instruction that caused the trap

Note now potentially 2 page transfers for page fault – increasing Effective
memory access time
 Page Replacement

5
5 6 5
6

6
 Page Replacement

5 5
6

6
 Page Replacement
• Modify (dirty) bit to reduce overhead of page
transfers
– only modified pages are written to disk
 Performance of Demand Paging
Demand paging affects the performance of the computer systems

• Page Fault Rate 0  p  1

– if p = 0 no page faults
– if p = 1, every reference is a fault

• Effective Access Time (EAT)

EAT = (1 – p) x memory access
+ p (page fault overhead
+ swap page out
+ swap page in
+ restart overhead
)
 Demand Paging Example
• Memory access time = 200 nanoseconds
• Average page-fault service time = 8 milliseconds

• EAT = (1 – p) x 200 + p (8 milliseconds)

= (1 – p ) x 200 + p x 8,000,000
= 200 + p x 7,999,800
• If one access out of 1,000 causes a page fault, then
EAT = 8.2 microseconds.
This is a slowdown by a factor of 40!!
• If want performance degradation < 10 percent
– 220 > 200 + 7,999,800 x p
20 > 7,999,800 x p
– p < .0000025
– < one page fault in every 400,000 memory accesses

Better utilization of swap space

Swap space
 Page replacement algorithm
Evaluation
• Evaluate algorithm by running it on a particular string
of memory references (reference string) and
computing the number of page faults on that string
– String is just page numbers, not full addresses
– Repeated access to the same page does not cause a page
fault
• Trace the memory reference of a process
0100, 0432, 0101, 0612, 0102, 0104, 0101, 0611, 0102
• Page size 100B
• Reference string 1, 4, 1, 6, 1, 6

• In all our examples, the reference string is

7,0,1,2,0,3,0,4,2,3,0,3,0,3,2,1,2,0,1,7,0,1
Graph of Page Faults Versus
The Number of Frames
 First-In-First-Out (FIFO) Algorithm

Associates a time with each frame when the page was brought
into memory

When a page must be replaced, the oldest one is chosen

Reference string: 7,0,1,2,0,3,0,4,2,3,0,3,0,3,2,1,2,0,1,7,0,1

Limitation:
A variable is initialized early and constantly used
 FIFO Page Replacement
3 frames (3 pages can be in memory at a time)

15 page faults

• How to track ages of pages?

• Just use a FIFO queue to hold all the pages in memory
• Replace the page at the head
• Insert at tail
 Optimal Algorithm

• Replace page that will not be used for longest

period of time

• How do you know this?

– Can’t read the future

• Used for measuring how well your algorithm

performs
Optimal Page Replacement

9 page faults
 Least Recently Used (LRU)
Algorithm
• Use past knowledge rather than future
– Past is the proxy of future
• Replace page that has not been used in the most of the
time
• Associate time of last use with each page

• 12 faults – better than FIFO but worse than OPT

• Generally good algorithm and frequently used
• But how to implement?
 LRU Algorithm-Implementation
• Counter implementation
– CPU maintains a clock
– Every page entry has a Time of use;
• every time page is referenced, copy the clock into the time of
use
– When a page needs to be replaced, look at the “Time of use” to
find smallest value
• Search through table needed
Page table

Time of
Clock Use
CPU
0
1
address
2
3
 LRU Algorithm-Implementation
• Stack implementation
• Keep a stack of page numbers in a double linked list form:
• Page referenced:
• move it to the top
• Victim page is the bottom page
 LRU Approximation Algorithms
• LRU needs special hardware and still slow
• Reference bit
– With each page associate a bit, initially = 0
– When page is referenced bit set to 1
• Additional reference bit algorithm
– Record the reference bits in regular interval
– Keep a 8 bit string for each page in memory
– At regular interval, timer copies the reference bit
to the high order bit (MSB) of the string.
– Shift the other bits right side by one bit
Reset
reference
LRU Approximation Algorithms
Period Period Period

0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5
 LRU Approximation Algorithms
• LRU needs special hardware and still slow
• Reference bit
– With each page associate a bit, initially = 0
– When page is referenced bit set to 1
• Additional reference bit algorithm
• Second-chance algorithm
– Generally FIFO, plus hardware-provided reference bit
– Clock replacement
– If page to be replaced has
• Reference bit = 0 -> replace it
• reference bit = 1 then:
– set reference bit 0, leave page in memory (reset the time)
– replace next page, subject to same rules
 Second-Chance (clock) Page-Replacement Algorithm

{Reference bit, Dirty bit} combination

 Counting Algorithms
• Keep a counter of the number of references
that have been made to each page
Least and most frequently used

• LFU Algorithm: replaces page with smallest

count

• MFU Algorithm: based on the argument that

the page with the smallest count was
probably just brought in and has yet to be
used
Graph of Page Faults Versus
The Number of Frames
Increase in page frame decreases page fault rate?

FIFO Example –
Program - 5 pages,
3 frames of Memory allocated
1 2 3 4 1 2 5 1 2 3 4 5

9 1 1 1 4 4 4 5 5 5 5 5 5
faults 2 2 2 1 1 1 1 1 3 3 3
3 3 3 2 2 2 2 2 4 4
 FIFO Example –
Program - 5 pages,
4 frames of Memory allocated
1 2 3 4 1 2 5 1 2 3 4 5

1 1 1 4 4 4 5 5 5 5 5 5
2 2 2 1 1 1 1 1 3 3 3
3 3 3 2 2 2 2 2 4 4

10 1 1 1 1 1 1 5 5 5 5 4 4
faults 2 2 2 2 2 2 1 1 1 1 5
3 3 3 3 3 3 2 2 2 2
4 4 4 4 4 4 3 3 3
 Belady's Anomaly
# of Page Faults

Number of Frames
cs431-cotter 126
 Belady’s Anomaly
• This most unexpected result is known as
Belady’s anomaly – for some page-
replacement algorithms, the page fault rate
may increase as the number of allocated
frames increases

• Is there a characterization of algorithms

susceptible to Belady’s anomaly?
 Stack Algorithms
• Certain page replacement algorithms are more “well behaved”
than others

• There are a set of paging algorithms whereby the set of pages

loaded with an allocation of m frames is always a subset of the
set of pages loaded with an allocation of m +1 frames. This
property is called the inclusion property

• Algorithms that satisfy the inclusion property are not subject

to Belady’s anomaly. FIFO does not satisfy the inclusion
property and is not a stack algorithm
• (In the FIFO example), the problem arises because the set of
pages loaded with a memory allocation of 3 frames is not
necessarily also loaded with a memory allocation of 4 frames
 LRU algorithm
6 5 3 6 4 3 9
3 6 4 3 9
In memory
3 frames 5 3 6 4 3
6 5 3 6 4
5 5 6
5 Outside
memory

3 6 4 3 9
5 3 6 4 3 In memory
4 frames 6 5 3 4
6
5 5 6
5
 Page-Buffering
• Systems commonly keep a pool of free frames.
• When a page fault occurs, a victim frame is chosen as before.
• However, the desired page is read into a free frame from the
pool before the victim is swapped out.
• This procedure allows the process to restart as soon as
possible, without waiting
• When the victim is later swapped out, its frame is added to
the free-frame pool.
 Page-Buffering
• Whenever the paging device is idle, a
modified page is selected and is written to
secondary storage.
• Its modify bit is then reset.
• This scheme increases the probability that
a page will be clean when it is selected for
replacement and will not need to be swapped
out.
 Page-Buffering
• Another modification is to keep a pool of free
frames
– but to remember which page was in each frame.
• Since the frame contents are not modified when
a frame is swapped to secondary storage,
– the old page can be reused directly from the free-
frame pool if it is needed before that frame is reused.
– No I/O is needed
• When a page fault occurs, we first check whether
the desired page is in the free-frame pool.
 Allocation of Frames
• How do we allocate the fixed amount of
memory among various processes?

• Single user system

– Trivial
 Allocation of Frames
• Each process needs minimum number of frames
• Minimum number is defined by the instruction set
• Page fault forces to restart the instruction
– Enough frames to hold all the pages for that instruction
• Example:
– Single address instruction (2 frames)
– Two address instruction (3 frames)
– Indirect Addressing mode
• Maximum of course is total frames in the system
• Two major allocation schemes
– fixed allocation
– proportional allocation
 Fixed and proportional Allocation
• Equal allocation – m frames and n processes
– Each process gets m/n
• For example, if there are 100 frames (after allocating frames
for the OS) and 5 processes, give each process 20 frames
– Keep some as free frame buffer pool
• Unfair for small and large sized processes
• Proportional allocation – Allocate according to the size of
process
– Dynamic as degree of multiprogramming, process sizes
change m = 64
si = size of process pi s1 = 10
s2 = 127
S =  si
10
m = total number of frames a1 =  64  5
137
si 127
ai = allocation for pi = m a2 =  64  59
S 137
 Priority Allocation
Allocation of frames
• Depends on multiprogramming level

• Use a proportional allocation scheme using

priorities along with size
 Global vs. Local Allocation
• Frames are allocated to various processes

• If process Pi generates a page fault

– select for replacement one of its frames
– select for replacement a frame from another process

• Local replacement – each process selects from only its own set of
allocated frames
– More consistent per-process performance
– But possibly under-utilized memory

• Global replacement – process selects a replacement frame from the

set of all frames; one process can take a frame from another
– But then process execution time can vary greatly
– But greater throughput ----- so more common
• Processes can not control its own page fault rate
– Depends on the paging behavior of other processes
 Thrashing
• If a process uses a set of “active pages”
– Number of allocated frames is less than that
• Page-fault
– Replace some “active” page
– But quickly need replaced “active” frame back
– Quickly a page fault, again and again
– Thrashing  a process is busy swapping pages in and out

• OS monitors CPU utilization

– If low? Increase the degree of multiprogramming
 Thrashing
• Global page replacement
– Process enters new phase (subroutine call) execution
– Page fault
– Taking frames from other processes
• Replace “active” frames of other processes
– These processes start page fault
– These faulting processes wait on the device queue for disk
• Ready queue empty
– CPU utilization decreases Disk

• CPU scheduler increases the degree of multiprogramming

– More page faults
– Drop in CPU utilization
• Page fault increases tremendously
Thrashing (Cont.)
 Thrashing
• Solution
– Local replacement
– One process cannot steal frames from other
processes
• Provide a process as many frames as needed
– Able to load all active pages
– How do we know?
– Locality model
 Demand Paging and Thrashing

• Why does demand paging work?

Locality model
– Process migrates from one locality to another
– Localities may overlap

• Allocate enough frames to a process to accommodate its

locality

• Why does thrashing occur?

 size of locality > total allocated frames
– Limit effects by using local or priority page replacement
Locality In A Memory-Reference Pattern
Working-set model
 Working-Set Model
•   working-set window  a fixed number of page references
Example: 10,000 instructions
• WSSi (working set of Process Pi) =
total number of pages referenced in the most recent  (varies
in time)
– if  too small will not encompass entire locality
– if  too large will encompass several localities
– if  =   will encompass entire program
• D =  WSSi  total demand frames
– Approximation of locality
• if D > m  Thrashing
• Policy if D > m, then suspend or swap out one of the
processes
Implementation of WS model
 Page-Fault Frequency
• More direct approach than WSS
• Establish “acceptable” page-fault frequency (PFF)
rate and use local replacement policy
– If actual rate too low, process loses frame
– If actual rate too high, process gains frame
 Problem
A certain computer system has the segmented paging
architecture for virtual memory. The memory is byte
addressable. Both virtual and physical address spaces
contain 216 bytes each. The virtual address space is
divided into 8 non-overlapping equal size segments. The
memory management unit (MMU) has a hardware
segment table, each entry of which contains the physical
address of the page table for the segment. Page tables
are stored in the main memory and consists of 2 byte
page table entries. What is the minimum page size in
bytes so that the page table for a segment requires at
most one page to store it?
• Let page size = n bytes.
• Now, since page table has to be stored into a single page, so we must have-
• Size of page table <= Page size
• Size of Each Segment-
• Size of each segment
= Process size / Number of segments
= 216 bytes / 8
= 216 bytes / 23
= 213 bytes
= 8 KB

Number of Pages of Each Segment-

• Number of pages each segment is divided
• = Size of segment / Page size
• = 8 KB / n bytes
• = (8K / n) pages
Size of Each Page Table-

Size of each page table

• = Number of entries in page table x Page table entry size
• = Number of pages the segment is divided x 2 bytes
• = (8K / n) x 2 bytes
• = (16K / n) bytes

Page Size

• Substituting values in the above condition, we get-

• (16K / n) bytes <= n bytes
• (16K / n) <= n
• n2 >= 16K
• n2 >= 214
• n >= 27

• Thus, minimum page size possible = 27 bytes = 128 bytes.