9781439079201_PPT_ch02

8/18/2019 9781439079201_PPT_ch02

1/69

8/18/2019 9781439079201_PPT_ch02

2/69

After completing this chapter, you shouldbe able to describe:

The basic functionality of the three

memory allocation schemes presented inthis chapter: xed partitions, dynamicpartitions, relocatable dynamic partitions

Best-t memory allocation as well asrst-t memory allocation schemes

How a memory list keeps track ofaailable memory

Understanding Operating Systems, Sixth

Edition 2!

8/18/2019 9781439079201_PPT_ch02

3/69

Understanding Operating Systems, SixthEdition 3

The importance of deallocation ofmemory in a dynamic partition system

The importance of the bounds register in

memory allocation schemes The role of compaction and how it

improes memory allocation e"ciency

#

8/18/2019 9781439079201_PPT_ch02

4/69

$anagement of main memory is critical% The performance of the entire system has

been directly dependent on two things:

How much memory is aailable How it is optimi&ation while 'obs are being

processed%

This chapter introduces: The memory manager ()A$* +ore memory (primary storage*


8/18/2019 9781439079201_PPT_ch02

5/69

This chapter introduces: our types of memory allocation schemes

.ingle-user systems

ixed partitions /ynamic partitions

)elocatable dynamic partitions

These early memory management schemesare seldom used by today0s 1.s but are

important to study because each oneintroduced fundamental concepts that helpedmemory management eole%


2

8/18/2019 9781439079201_PPT_ch02

6/69

+ommercially aailable in 345s and 3425s

6ach program was loaded in its entirety intomemory and allocated as much contiguous

memory space allocated as it needed% 7f the program was too large and didn0t t the

aailable memory space, it couldn0t be executed,

Although early computers were physically

large, they had ery little memory% +omputers hae only a nite amount of

memory%


8/18/2019 9781439079201_PPT_ch02

7/69

7f a program doesn0t t, then either thesi&e of the main memory must beincreased or the program must be

modied% $aking it smaller

8sing methods that allow program segments(partitions made to the program* to beoerlaid%

Transfer segments of a program from secondarystorage into main memory for execution

Two or more segments take turns occupying thesame memory locations%


8/18/2019 9781439079201_PPT_ch02

8/69

9rogram .egmentation .lides


8/18/2019 9781439079201_PPT_ch02

9/69

8/18/2019 9781439079201_PPT_ch02

10/69

8/18/2019 9781439079201_PPT_ch02

11/69

M o d u l e 1 : M a i n c o n t r o l a n d k e y d a t a

M o d u l e 2 : N o r m a l d a t a p r o c e s s i n g l o g i c

M o d u l e 3 : E r r o r p r o c e s s i n g

M o d u l e 4 : E n d - o f - j o b s u m m a r y c o m p u t a t i o n s

8/18/2019 9781439079201_PPT_ch02

12/69


M o d u l e 2 : N o r m a l d a t a p r o c e s s i n g l o g i c

8/18/2019 9781439079201_PPT_ch02

13/69


M o d u l e 3 : E r r o r p r o c e s s i n g

8/18/2019 9781439079201_PPT_ch02

14/69


M o d u l e 4 : E n d - o f - j o b s u m m a r y c o m p u t a t i o n s

8/18/2019 9781439079201_PPT_ch02

15/69

A program cannot process data it doesnot hae%

A program spends more time waiting for

71 than processing data%

8/18/2019 9781439079201_PPT_ch02

16/69

The amount of work performed by the$emory $anager is minimal%

1nly two hardware items are needed:

A register to store the base address An accumulator to keep track of the si&e of

the program as it0s being read into memory%

1nce the program is entirely loaded into

memory, it remains there until executionis complete, either through normaltermination or by interention of the 1.%


8/18/2019 9781439079201_PPT_ch02

17/69

/isadantages There is no support for multiprogramming or

networking

7t can handle only one 'ob at a time% This conguration was rst used in research

institutions but proed unacceptable for thebusiness community%

7t was not cost e;ectie to spend almost


8/18/2019 9781439079201_PPT_ch02

18/69

The rst attempt to followmultiprogramming used fxed partitions (static partitions* within main memory%

1ne partition for each 'ob% The si&e of each partition was designated

when the system was powered on% 6ach partition could only be recongured

when the system was shut down% 1nce the system was in operation, thepartition si&es remained static%


8/18/2019 9781439079201_PPT_ch02

19/69

7ntroduced protection of the 'ob0s memoryspace 1nce a partition was assigned to a 'ob, no other 'ob

could be allowed to enter its boundaries, eitheraccidentally or intentionally% >ot a problem in single-user contiguous allocation

schemes%

ixed 9artitions is more ?exible than the .ingle-8ser scheme because it allows seeralprograms to be in memory at the same time%

ixed 9artitions still re=uires that the entireprogram be stored contiguously and in memoryfrom the beginning to the end of its execution%


8/18/2019 9781439079201_PPT_ch02

20/69

7n order to allocate memory spaces to 'obs, the 1.0s $emory $anager mustkeep a table which shows: 6ach memory partition si&e 7ts address 7ts access restrictions

7ts current status (free or busy*

As each 'ob terminates, the status of itsmemory partition is changed from busyto free so an incoming 'ob can beassigned to that partition%


8/18/2019 9781439079201_PPT_ch02

21/69

$emory manager allocates memory space to 'obs 8ses a table


8/18/2019 9781439079201_PPT_ch02

22/69


8/18/2019 9781439079201_PPT_ch02

23/69

/isadantages The ixed 9artition scheme works well if all the 'obs run on the system are of the same si&e orif the si&es are known ahead of time and don0t

ary between reconguration% 7f the partition si&es are too small:

larger 'obs will be re'ected if they0re too big to tinto the largest partition%

@arge 'obs will wait if the large partitions arebusy%

@arge 'obs may hae a longer turnaround time asthey wait for free partitions of su"cient si&e ormay neer run%


8/18/2019 9781439079201_PPT_ch02

24/69

/isadantages 7f the partition si&es are too big, memory is

wasted% 7f a 'ob does not occupy the entire partition, the

unused memory in the partition will remain idle% 7t can0t be gien to another 'ob because each

partition is allocated to only one 'ob at a time%

9artial usage of xed partitions and the

coinciding creation of unused spaces withinthe partition is called internalfragmentation and is a ma'or drawback tothe xed partition memory allocation scheme%


8/18/2019 9781439079201_PPT_ch02

25/69

ith Dynamic Partitions aailable memoryis still kept in contiguous blocks but 'obs aregien only as much memory as they re=uest

when they are loaded for processing% A dynamic partition scheme fully utili&es

memory when the rst 'obs are loaded%

As new 'obs enter the system that are not the

same si&e as those that 'ust acated memory,they are t into the aailable spaces on apriority basis% irst +ome irst .ere priority


8/18/2019 9781439079201_PPT_ch02

26/69

The subse=uent allocation of memory createsfragments of free memory between blocks ofallocated memory%

6xternal fragmentation

@ets memory go to waste


8/18/2019 9781439079201_PPT_ch02

27/69


8/18/2019 9781439079201_PPT_ch02

28/69

or both xed and dynamic memoryallocation schemes, the 1. must keeplists of each memory location notingwhich are free and which are busy%

As new 'obs come into the system, thefree partitions must be allocated%

These partitions may be allocated on thebasis of: First-t memory allocation:

irst partition tting the re=uirements

Best-t memory allocation:

.mallest partition tting the re=uirements

@east wasted spaceUnderstanding Operating Systems, SixthEdition 28

8/18/2019 9781439079201_PPT_ch02

29/69

or both schemes, the $emory $anagerorgani&es the memory lists of the freeand used partitions either by si&e or by

location% The Best-t allocation method keeps the

freebusy lists in order by si&e, smallest tolargest%

$akes the best use of memory space%

The rst-t method keeps the freebust listorgani&ed by memory locations, low-ordermemory to high-order memory%

aster in making the allocation%


8/18/2019 9781439079201_PPT_ch02

30/69


8/18/2019 9781439079201_PPT_ch02

31/69


8/18/2019 9781439079201_PPT_ch02

32/69

Algorithm for rst-t Assumes memory manager keeps two lists

1ne for free memory blocks

1ne for busy memory blocks

The operation consists of a simple loop thatcompares the si&e of each 'ob to the si&e ofeach memory block until a block is found thatis large enough to t the 'ob%

The 'ob is stored into that block of memoryand the $emory $anager moes out of theloop to fetch the next 'ob from the entry =ueue


8/18/2019 9781439079201_PPT_ch02

33/69

Algorithm for rst-t (cont'd.): 7f the entire list is searched in ain, then the 'ob is placed into a waiting =ueue%

The $emory $anager then fetches the next 'ob and repeats the process%


8/18/2019 9781439079201_PPT_ch02

34/69


8/18/2019 9781439079201_PPT_ch02

35/69

Algorithm for best-t The goal is to nd the smallest memory block

into which the 'ob will t

The entire table must be searched before theallocation can be made because the memoryblocks are physically stored in se=uenceaccording to their location in memory%


8/18/2019 9781439079201_PPT_ch02

36/69


8/18/2019 9781439079201_PPT_ch02

37/69

Hypothetical allocation schemes >ext-t:

.tarts searching from last allocated block, fornext aailable block when a new 'ob arries

orst-t:

Allocates largest free aailable block to new 'ob

1pposite of best-t

Cood way to explore theory of memoryallocation

>ot best choice for an actual system


8/18/2019 9781439079201_PPT_ch02

38/69

There eentually comes a time whenmemory space must be released ordeallocated %

or a xed-partition system: hen the 'ob is completed, the $emory

$anager resets the status of the memoryblock where the 'ob was stored to DfreeE%

Any code may be used% 6xample code: binary alues with &ero indicating

free and one indicating busy%


8/18/2019 9781439079201_PPT_ch02

39/69

or dynamic-partition system: A more complex Algorithm is used because

the algorithm tries to combine free areas ofmemory wheneer possible%

The system must be prepared for threealternatie solutions:

Case : hen the block to be deallocated isadjacent to another free block

Case !: hen the block to be deallocated isbetween two free blocks

Case ": hen the block to be deallocated isisolated from other free blocks


8/18/2019 9781439079201_PPT_ch02

40/69

8sing the deallocation algorithm, the systemsees that the memory to be released is next toa free memory block (@ocation FG55*%

The @ist must be changed changes to re?ect

the starting address of the new free block(F55* which was the address of the rstinstruction of the 'ob that 'ust released thisblock%

7n addition, the memory block si&e for this newfree space must be changed to show its newsi&e, which is the combined total of the two freepartitions (!55 I 2*%


8/18/2019 9781439079201_PPT_ch02

41/69


8/18/2019 9781439079201_PPT_ch02

42/69


8/18/2019 9781439079201_PPT_ch02

43/69

hen the deallocated memory space isbetween two free memory blocks:

8sing the deallocation algorithm, the systemlearns that the memory to be deallocated isbetween two free blocks of memory%

The si&es of the three free partitions (!5 I !5I!52* must be combined and the total stored

with the smallest beginning address, F25% Because the entry at location F55 has been

combined with the preious entry, we mustempty out this entry%


8/18/2019 9781439079201_PPT_ch02

44/69

e do that by changing the status to n#llentry, with no beginning address and nomemory block si&e as indicated by an asterisk%

This negates the need to rearrange the list at

the expense of memory%


8/18/2019 9781439079201_PPT_ch02

45/69


8/18/2019 9781439079201_PPT_ch02

46/69


8/18/2019 9781439079201_PPT_ch02

47/69

The third alternatie is when the space to bedeallocated is isolated from all other freeareas% 8sing the deallocation algorithm, the system

learns that the memory block to be released isnot ad'acent to any free blocks of memory

7t is between two other busy areas

The system must search the table for a null entry%

hen the null entry is found:

The beginning memory location of the terminating 'ob is entered in the beginning address column%


8/18/2019 9781439079201_PPT_ch02

48/69

The 'ob si&e is entered under the memory blocksi&e column

The status is changed from a null entry to free toindicate that a new block of memory is aailable%


8/18/2019 9781439079201_PPT_ch02

49/69


8/18/2019 9781439079201_PPT_ch02

50/69


8/18/2019 9781439079201_PPT_ch02

51/69


8/18/2019 9781439079201_PPT_ch02

52/69

8/18/2019 9781439079201_PPT_ch02

53/69

Both the xed and dynamic memoryallocation schemes described thus farshared some unacceptable fragmentationcharacteristics that had to be resoledbefore the number of 'obs waiting to beaccepted became unwieldy%

7n addition, there was a growing need to useall the sliers of memory often left oer%

The solution to both problems was thedeelopment of relocatable dynamic partitions%


8/18/2019 9781439079201_PPT_ch02

54/69

Relocatable Dynamic Partitions The $emory $anager relocates programs to

gather together all of the empty blocks andcompact them to make one block of memory

large enough to accommodate some or all ofthe 'obs waiting to get in%

The compaction of memory is performedby the 1. to reclaim fragmented sections

of the memory space% 6ery program in memory must be relocated so

they0re contiguous%


8/18/2019 9781439079201_PPT_ch02

55/69

$he compaction of memory 6ery address, and eery reference to an

address, within each program must bead'usted to account for the program0s new

location in memory% All other alues within the program must be

left alone%


8/18/2019 9781439079201_PPT_ch02

56/69


8/18/2019 9781439079201_PPT_ch02

57/69


8/18/2019 9781439079201_PPT_ch02

58/69


8/18/2019 9781439079201_PPT_ch02

59/69


8/18/2019 9781439079201_PPT_ch02

60/69

8/18/2019 9781439079201_PPT_ch02

61/69

%hat lists ha&e to be #pdated After relocation and compaction, both the free

list and the busy list are updated%

ree list is changed to show the partition for the

new block of free memory The one formed as a result of compaction that will

be located in memory starting after the last locationused by the last 'ob%

Busy list is changed to show the new locationsfor all of the 'obs already in process that wererelocated%


8/18/2019 9781439079201_PPT_ch02

62/69

.pecial-purpose registers used for relocation: Bo#nds register

8sed to store the highest (or lowest* location inmemory accessible by each program%

This ensures that, during execution, a programwon0t try to access memory locations that don0tbelong to it%

elocation register

+ontains the alue that must be added to each

address referenced in the program so that thesystem will be able to access the correct memoryaddresses after relocation%

7f the program isn0t relocated, the alue, D&eroE isstored in the program0s relocation register%


8/18/2019 9781439079201_PPT_ch02

63/69

By compacting and relocating, the$emory $anager optimi&es use ofmemory and, thus, 7mproes throughput%

An unfortunate side e;ect is that moreoerhead is incurred than with the twopreious memory allocation schemes%

The crucial factor is the timing of the

compaction% hen and how often should it be done%


8/18/2019 9781439079201_PPT_ch02

64/69

There are three options for the timing ofcompaction: 1ne approach is to do it when a certain

percentage of memory becomes busy F2K

The disadantage of this approach is that thesystem would incur unnecessary oerhead if no 'obswere waiting to use the remaining !2K%

A second approach is to compact memory onlywhen there are 'obs waiting to get in%

This would entail constant checking of the entry=ueue, which might result in unnecessary oerheadand slow down the processing of the 'obs in thesystem%


8/18/2019 9781439079201_PPT_ch02

65/69

There are three options for the timing ofcompaction: A third approach is to do it after a prescribed

amount of time has elapsed%

7f the amount of time chosen is too small, thesystem will spend more time on compaction thanon processing%

7f the amount of time is too large, too many 'obs

will congregate in the waiting =ueue and theadantages of compaction are lost%


8/18/2019 9781439079201_PPT_ch02

66/69

There are three options for the timing ofcompaction: The best choice for any system is decided by

the 1. designer who, based on the 'ob mix

and other factors, tries to optimi&e bothprocessing time and memory use whilekeeping oerhead as low as possible%


8/18/2019 9781439079201_PPT_ch02

67/69

our memory management techni=ues .ingle-user systems

ixed partitions

/ynamic partitions )elocatable dynamic partitions

+ommon re=uirements of four memorymanagement techni=ues 6ntire program loaded into memory

+ontiguous storage

$emory residency until 'ob completed


8/18/2019 9781439079201_PPT_ch02

68/69

6ach places seere restrictions on 'obsi&e

.u"cient for rst three generations of

computers which processed 'obs in batchmode% Turnaround time was measured in hours%

Understanding Operating Systems, Sixth

Edition 68

8/18/2019 9781439079201_PPT_ch02

69/69

>ew modern memory managementtrends in late 345s and early 34F5s /iscussed in next chapter

+ommon characteristics of memory schemes 9rograms are not stored in contiguous memory

>ot all segments reside in memory during 'obexecution

9781439079201_PPT_ch02

Documents