LECTURER NOTES MODULE 4: MEMORY MANAGMENT

Module 4: Memory Management
The von Neumann principle for the design and operation of computers requires that a
program has to be primary memory resident to execute. Also, a user requires to revisit his
programs often during its evolution. However, due to the fact that primary memory is
volatile, a user needs to store his program in some non-volatile store. All computers
provide a non-volatile secondary memory available as an online storage. Programs and
files may be disk resident and downloaded whenever their execution is required.
Therefore, some form of memory management is needed at both primary and secondary
memory levels.
Secondary memory may store program scripts, executable process images and data files.
It may store applications, as well as, system programs. In fact, a good part of all OS, the
system programs which provide services (the utilities for instance) are stored in the
secondary memory. These are requisitioned as needed.
The main motivation for management of main memory comes from the support for multiprogramming.
Several executables processes reside in main memory at any given time. In
other words, there are several programs using the main memory as their address space.
Also, programs move into, and out of, the main memory as they terminate, or get
suspended for some IO, or new executables are required to be loaded in main memory.
So, the OS has to have some strategy for main memory management. In this chapter we
shall discuss the management issues and strategies for both main memory and secondary
memory.
4.1 Main Memory Management
Let us begin by examining the issues that prompt the main memory management.
􀂾 Allocation: First of all the processes that are scheduled to run must be resident in
the memory. These processes must be allocated space in main memory.
􀂾 Swapping, fragmentation and compaction: If a program is moved out or
terminates, it creates a hole, (i.e. a contiguous unused area) in main memory.
When a new process is to be moved in, it may be allocated one of the available
holes. It is quite possible that main memory has far too many small holes at a
certain time. In such a situation none of these holes is really large enough to be
allocated to a new process that may be moving in. The main memory is too
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/2
fragmented. It is, therefore, essential to attempt compaction. Compaction means
OS re-allocates the existing programs in contiguous regions and creates a large
enough free area for allocation to a new process.
􀂾 Garbage collection: Some programs use dynamic data structures. These
programs dynamically use and discard memory space. Technically, the deleted
data items (from a dynamic data structure) release memory locations. However, in
practice the OS does not collect such free space immediately for allocation. This
is because that affects performance. Such areas, therefore, are called garbage.
When such garbage exceeds a certain threshold, the OS would not have enough
memory available for any further allocation. This entails compaction (or garbage
collection), without severely affecting performance.
􀂾 Protection: With many programs residing in main memory it can happen that due
to a programming error (or with malice) some process writes into data or
instruction area of some other process. The OS ensures that each process accesses
only to its own allocated area, i.e. each process is protected from other processes.
􀂾 Virtual memory: Often a processor sees a large logical storage space (a virtual
storage space) though the actual main memory may not be that large. So some
facility needs to be provided to translate a logical address available to a processor
into a physical address to access the desired data or instruction.
􀂾 IO support: Most of the block-oriented devices are recognized as specialized
files. Their buffers need to be managed within main memory alongside the other
processes. The considerations stated above motivate the study of main memory
management.
One of the important considerations in locating an executable program is that it should be
possible to relocate it any where in the main memory. We shall dwell upon the concept of
relocation next.
4.2 Memory Relocation Concept
Relocation is an important concept. To understand this concept we shall begin with a
linear map (one-dimensional view) of main memory. If we know an address we can fetch
its contents. So, a process residing in the main memory, we set the program counter to an
absolute address of its first instruction and can initiate its run. Also, if we know the
locations of data then we can fetch those too. All of this stipulates that we know the
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/3
Figure 4.1: The relocation concept.
absolute addresses for a program, its data and process context etc. This means that we can
load a process with only absolute addresses for instructions and data, only when those
specific addresses are free in main memory. This would mean we loose flexibility with
regard to loading a process. For instance, we cannot load a process, if some other process
is currently occupying that area which is needed by this process. This may happen even
though we may have enough space in the memory. To avoid such a catastrophe,
processes are generated to be relocatable. In Figure 4.1 we see a process resident in main
memory.
Initially, all the addresses in the process are relative to the start address. With this
flexibility we can allocate any area in the memory to load this process. Its instruction,
data, process context (process control block) and any other data structure required by the
process can be accessed easily if the addresses are relative. This is most helpful when
processes move in and out of main memory. Suppose a process created a hole on moving
out. In case we use non-relocatable addresses, we have the following very severe
problem.
When the process moves back in, that particular hole (or area) may not be available any
longer. In case we can relocate, moving a process back in creates no problem. This is so
because the process can be relocated in some other free area. We shall next examine the
linking and loading of programs to understand the process of relocation better.
4.2.1 Compiler Generated Bindings
The advantage of relocation can also be seen in the light of binding of addresses to
variables in a program. Suppose we have a program variable x in a program P. Suppose
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/4
the compiler allocated a fixed address to x. This address allocation by the compiler is
called binding. If x is bound to a fixed location then we can execute program P only when
x could be put in its allocated memory location. Otherwise, all address references to x
will be incorrect.
If, however, the variable can be assigned a location relative to an assumed origin (or first
address in program P) then, on relocating the program's origin anywhere in main
memory, we will still be able to generate a proper relative address reference for x and
execute the program. In fact, compilers generate relocatable code. In the next section we
describe the linking, loading, and relocation of object code in greater detail.
4.3 Linking and Loading Concepts
In Figure 4.2 we depict the three stages of the way a HLL program gets processed.
Figure 4.2: Linking and loading.
The three stages of the processing are:
􀂾 Stage 1: In the first stage the HLL source program is compiled and an object code
is produced. Technically, depending upon the program, this object code may by
itself be sufficient to generate a relocatable process. However many programs are
compiled in parts, so this object code may have to link up with other object
modules. At this stage the compiler may also insert stub at points where run time
library modules may be linked.
􀂾 Stage 2: All those object modules which have sufficient linking information
(generated by the compiler) for static linking are taken up for linking. The linking
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/5
editor generates a relocatable code. At this stage, however, we still do not replace
the stubs placed by compilers for a run time library link up.
􀂾 Stage3: The final step is to arrange to make substitution for the stubs with run
time library code which is a relocatable code.
When all the three stages are completed we have an executable. When this executable is
resident in the main memory it is a runnable process.
Recall our brief discussion in the previous section about the binding of variables in a
program. The compiler uses a symbol table to generate addresses. These addresses are
not bound, i.e. these do not have absolute values but do have information on sizes of data.
The binding produced at compile time is generally relative. Some OSs support a linking
loader which translates the relative addresses to relocatable addresses. In any event, the
relocatable process is finally formed as an output of a loader.
4.4 Process and Main Memory Management
Once processes have been created, the OS organizes their execution. This requires
interaction between process management and main memory management. To understand
this interaction better, we shall create a scenario requiring memory allocations. For the
operating environment we assume the following:
􀂾 A uni-processor, multi-programming operation.
􀂾 A Unix like operating system environment.
With a Unix like OS, we can assume that main memory is partitioned in two parts. One
part is for user processes and the other is for OS. We will assume that we have a main
memory of 20 units (for instance it could be 2 or 20 or 200 MB). We show the
requirements and time of arrival and processing requirements for 6 processes in Table
4.1.
Table 4.1: The given data.
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/6
We shall assume that OS requires 6 units of space. To be able to compare various
policies, we shall repeatedly use the data in Table 4.1 for every policy option. In the next
section we discuss the first fit policy option.
With these requirements we can now trace the emerging scenario for the given data. We
shall assume round robin allocation of processor time slots with no context switching
Table 4.2: FCFS memory allocation.
over-heads. We shall trace the events as they occur giving reference to the corresponding
part in Table 4.2. This table also shows a memory map as the processes move in and out
of the main memory.
4.5 The First Fit Policy: Memory Allocation
In this example we make use of a policy called first fit memory allocation policy. The
first fit policy suggests that we use the first available hole, which is large enough to
accommodate an incoming process. In Figure 4.3, it is important to note that we are
following first-come first-served (process management) and first fit (memory allocation)
policies. The process index denotes its place in the queue. As per first-come first-served
policies the queue order determines the order in which the processes are allocated areas.
In addition, as per first-fit policy allocation we scan the memory always from one end
and find the first block of free space which is large enough to accommodate the incoming
process.
In our example, initially, processes P1, P2, P3 and P4 are in the queue. The allocations
for processes P1, P2, P3 are shown in 4.3(a). At time 5, process P2 terminates. So,
process P4 is allocated in the hole created by process P2. This is shown at 4.3(b) in the
figure. It still leaves a hole of size 3. Now on advancing time further we see that at time
8, process P1 terminates. This creates a hole of size 3 as shown at 4.3(c) in the figure.
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/7
This hole too is now available for allocation. We have 3 holes at this stage. Two of these
3 holes are of size 3 and one is of size 2. When process P5 arrives at time 10, we look for
the first hole which can accommodate it. This is the one created by the departure of
process P1. Using the first-fit argument this is the hole allocated to process P5 as shown
in Figure 4.3(d). The final allocation status is shown in Figure 4.3. The first-fit allocation
policy is very easy to implement and is fast in execution.
Figure 4.3: First-fit policy allocation.
4.6 The Best Fit Policy: Memory Allocation
The main criticism of first-fit policy is that it may leave many smaller holes. For instance,
let us trace the allocation for process P5. It needs 2 units of space. At the time it moves
into the main memory there is a hole with 2 units of space. But this is the last hole when
we scan the main memory from the top (beginning). The first hole is 3 units. Using the
first-fit policy process P5 is allocated this hole. So when we used this hole we also
created a still smaller hole. Note that smaller holes are less useful for future allocations.
In the best-fit policy we scan the main memory for all the available holes. Once we have
information about all the holes in the memory then we choose the one which is closest to
the size of the requirement of the process. In our example we allocate the hole with size 2
as there is one available. Table 4.3 follows best-fit policy for the current example.
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/8
Also, as we did for the previous example, we shall again assume round-robin allocation
of the processor time slots. With these considerations we can now trace the possible
emerging scenario.
In Figure 4.4, we are following first-come first-served (process management) and best fit
(memory allocation) policies. The process index denotes its place in the queue. Initially,
processes P1, P2, P3 and P4 are in the queue. Processes P1, P2 and P3 are allocated as
Figure 4.4: Best-fit policy allocation
shown in Figure 4.4(a). At time 5, P2 terminates and process P4 is allocated in the hole so
created. This is shown in Figure 4.4(b). This is the best fit. It leaves a space of size 3
creating a new hole. At time 8, process P1 terminates. We now have 3 holes. Two of
these holes are of size 3 and one is of size 2. When process P5 arrives at time 10, we look
for a hole whose size is nearest to 2 and can accommodate P5. This is the last hole.
Clearly, the best-fit (and also the worst-fit) policy should be expected to be slow in
execution. This is so because the implementation requires a time consuming scan of all of
main memory. There is another method called the next-fit policy. In the next-fit method
the search pointer does not start at the top (beginning), instead it begins from where it
ended during the previous search. Like the first-fit policy it locates the next first-fit hole
that can be used. Note that unlike the first-fit policy the next-fit policy can be expected to
distribute small holes uniformly in the main memory. The first-fit policy would have a
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/9
tendency to create small holes towards the beginning of the main memory scan. Both
first-fit and next-fit methods are very fast and easy to implement.
In conclusion, first-fit and next-fit are the fastest and seem to be the preferred methods.
One of the important considerations in main memory management is: how should an OS
allocate a chunk of main memory required by a process. One simple approach would be
to somehow create partitions and then different processes could reside in different
partitions. We shall next discuss how the main memory partitions may be created.
Table 4.3: Best-fit policy memory allocation.
4.7 Fixed and Variable Partitions
In a fixed size partitioning of the main memory all partitions are of the same size. The
memory resident processes can be assigned to any of these partitions. Fixed sized
partitions are relatively simple to implement. However, there are two problems. This
scheme is not easy to use when a program requires more space than the partition size. In
this situation the programmer has to resort to overlays. Overlays involve moving data and
program segments in and out of memory essentially reusing the area in main memory.
The second problem has to do with internal fragmentation. No matter what the size of the
process is, a fixed size of memory block is allocated as shown in Figure 4.5(a). So there
will always be some space which will remain unutilized within the partition.
In a variable-sized partition, the memory is partitioned into partitions with different sizes.
Processes are loaded into the size nearest to its requirements. It is easy to always ensure
the best-fit. One may organize a queue for each size of the partition as shown in the
Figure 4.5(b). With best-fit policy, variable partitions minimize internal fragmentation.
However, such an allocation may be quite slow in execution. This is so because a process
may end up waiting (queued up) in the best-fit queue even while there is space available
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/10
elsewhere. For example, we may have several jobs queued up in a queue meant for jobs
that require 1 unit of memory, even while no jobs are queued up for jobs that require say
4 units of memory.
Both fixed and dynamic partitions suffer from external fragmentation whenever there are
partitions that have no process in it. One of techniques that have been used to keep both
internal and external fragmentations low is dynamic partitioning. It is basically a variable
partitioning with a variable number of partitions determined dynamically (i.e. at run
time).
Figure 4.5: Fixed and variable sized partitions.
Such a scheme is difficult to implement. Another scheme which falls between the fixed
and dynamic partitioning is a buddy system described next.
Figure 4.6: Buddy system allocation.
The Buddy system of partitioning: The buddy system of partitioning relies on the fact
that space allocations can be conveniently handled in sizes of power of 2. There are two
ways in which the buddy system allocates space. Suppose we have a hole which is the
closest power of two. In that case, that hole is used for allocation. In case we do not have
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/11
that situation then we look for the next power of 2 hole size, split it in two equal halves
and allocate one of these. Because we always split the holes in two equal sizes, the two
are \buddies". Hence, the name buddy system. We shall illustrate allocation using a
buddy system. We assume that initially we have a space of 1024 K. We also assume that
processes arrive and are allocated following a time sequence as shown in figure 4.6.
With 1024 K or (1 M) storage space we split it into buddies of 512 K, splitting one of
them to two 256 K buddies and so on till we get the right size. Also, we assume scan of
memory from the beginning. We always use the first hole which accommodates the
process. Otherwise, we split the next sized hole into buddies. Note that the buddy system
begins search for a hole as if we had a fixed number of holes of variable sizes but turns
into a dynamic partitioning scheme when we do not find the best-fit hole. The buddy
system has the advantage that it minimizes the internal fragmentation. However, it is not
popular because it is very slow. In Figure 4.6 we assume the requirements as (P1:80 K);
(P2:312 K); (P3:164 K); (P4:38 K). These processes arrive in the order of their index and
P1 and P3 finish at the same time.
4.8 Virtual Storage Space and Main Memory Partitions
Programming models assume the presence of main memory only. Therefore, ideally we
would like to have an unlimited (infinite) main memory available. In fact, an unlimited
main memory shall give us a Turing machine capability. However, in practice it is
infeasible. So the next best thing is attempted. CPU designers support and generate a very
large logical addressable space to support programming concerns. However, the directly
addressable main memory is limited and is quite small in comparison to the logical
addressable space. The actual size of main memory is referred as the physical memory.
The logical addressable space is referred to as virtual memory. The notion of virtual
memory is a bit of an illusion. The OS supports and makes this illusion possible. It does
so by copying chunks of disk memory into the main memory as shown in Figure 4.7. In
other words, the processor is fooled into believing that it is accessing a large addressable
space. Hence, the name virtual storage space. The disk area may map to the virtual space
requirements and even beyond.
Besides the obvious benefit that virtual memory offers a very large address space, there is
one other major benefit derived from the use of virtual storage. We now can have many
more main memory resident active processes. This can be explained as follows. During
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/12
much of the lifetime of its execution, a process operates on a small set of instructions
within a certain neighborhood. The same applies for the data as well. In other words a
process makes use of a very small memory area for doing most of the instructions and
making references to the data. As explained in Section 4.9, this is primarily due to the
locality of reference. So, technically, at any time we need a very small part of a process to
really be memory resident. For a moment, let us suppose that this small part is only
1/10th of the process's overall requirements. Note in that case, for the same size of
physical main memory, we can service 10 times as many memory resident programs. The
next question then is how do we organize and allocate these small chunks of often
required areas to be in memory. In fact, this is where paging and segmentation become
important. In this context we need to understand some of the techniques of partitioning of
main memory into pages or segments.
Figure 4.7: Virtual storage concept.
In addition, we need to understand virtual addressing concepts with paging and/or
segmentation. We begin with some simple techniques of partitioning both these
memories and management of processes.
4.9 Virtual Memory: Paging
In some sense, paging of virtual memory has an underlying mechanism which resembles
reading of a book. When we read a book we only need to open only the current page to
read. All the other pages are not visible to us. In the same manner, we can argue that even
when we may have a large online main memory available, the processor only needs a
small set of instructions to execute at any time. In fact, it often happens that for a brief
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/13
while, all the instructions which the processor needs to execute are within a small
proximity of each other. That is like a page we are currently reading in a book. Clearly,
this kind of situation happens quite frequently.
Essentially virtual memory is a large addressable space supported by address generating
mechanisms in modern CPUs. Virtual address space is much larger than the physical
main memory in a computer system. During its execution, a process mostly generates
instruction and data references from within a small range. This is referred to as the
locality of reference. Examples of locality of reference abound. For instance, we have
locality of reference during execution of a for or while loop, or a call to a procedure.
Even in a sequence of assignment statements, the references to instructions and data are
usually within a very small range. Which means, during bursts of process execution, only
small parts of all of the instruction and data space are needed, i.e. only these parts need be
in the main memory. The remaining process, instructions and data, can be anywhere in
the virtual space (i.e. it must remain accessible by CPU but not necessarily in main
memory). If we are able to achieve that, then we can actually follow a schedule, in which
we support a large address space and keep bringing in that part of process which is
needed. This way we can comfortably support (a) multi-programming (b) a large logical
addressable space giving enormous freedom to a programmer. Note, however, that this
entails mapping of logical addresses into physical address space. Such a mapping assures
that the instruction in sequence is fetched or the data required in computation is correctly
used.
If this translation were to be done in software, it would be very slow. In fact, nowadays
this address translation support is provided by hardware in CPUs. Paging is one of the
popular memory management schemes to implement such virtual memory management
schemes. OS software and the hardware address translation between them achieve this.
4.9.1 Mapping the Pages
Paging stipulates that main memory is partitioned into frames of sufficiently small sizes.
Also, we require that the virtual space is divided into pages of the same size as the
frames. This equality facilitates movement of a page from anywhere in the virtual space
(on disks) to a frame anywhere in the physical memory. The capability to map “any
page" to “any frame" gives a lot of flexibility of operation as shown in Figure 4.8
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/14
Division of main memory into frames is like fixed partitioning. So keeping the frame size
small helps to keep the internal fragmentation small. Often, the page to frame movement
is determined by a convenient size (usually a power of two) which disks also use for their
own DMA data transfer. The usual frame size is 1024 bytes, though it is not unusual to
have 4 K frame sizes as well. Paging supports multi-programming. In general there can
be many processes in main memory, each with a different number of pages. To that
extent, paging is like dynamic variable partitioning.
Figure 4.8: Paging implementation.
4.10 Paging: Implementation
Paging implementation requires CPU (HW) and OS (SW) support. In Figure 4.8, we
assume presence of three active processes. These processes need to have their pages
mapped to the main memory page frames. The OS maintains a page table for every
process to translate its logical to physical addresses. The page table may itself be resident
in main memory.
For a process, which is presently active, there are a number of pages that are in the main
memory. This set of pages (being used by the process) forms its resident set. With the
locality of reference generally observed, most of the time, the processes make reference
within the resident set. We define the set of pages needed by a process at any time as the
working set. The OS makes every effort to have the resident set to be the same as the
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/15
working set. However, it does happen (and happens quite often), that a page required for
continuing the process is not in the resident set. This is called a page fault. In normal
course of operation, though whenever a process makes virtual address reference, its page
table is looked up to find if that page is in main memory. Often it is there. Let us now
suppose that the page is not in main memory, i.e. a page fault has occurred. In that case,
the OS accesses the required page on the disk and loads it in a free page frame. It then
makes an entry for this page in process page table. Similarly, when a page is swapped
out, the OS deletes its entry from the page table. Sometimes it may well happen that all
the page frames in main memory are in use. If a process now needs a page which is not in
main memory, then a page must be forced out to make way for the new page. This is
done using a page replacement policy discussed next.
4.11 Paging: Replacement
Page replacement policies are based on the way the processes use page frames. In our
example shown in Figure 4.9, process P29 has all its pages present in main memory.
Process P6 does not have all its pages in main memory. If a page is present we record
1 against its entry. The OS also records if a page has been referenced to read or to write.
In both these cases a reference is recorded. If a page frame is written into, then a
modified bit is set. In our example frames 4, 9, 40, 77, 79 have been referenced and page
frames 9 and 13 have been modified. Sometimes OS may also have some information
about protection using rwe information. If a reference is made to a certain virtual address
. .
Figure 4.9: Replacement policy.
and its corresponding page is not present in main memory, then we say a page fault has
occurred. Typically, a page fault is followed by moving in a page. However, this may
require that we move a page out to create a space for it. Usually this is done by using an
appropriate page replacement policy to ensure that the throughput of a system does not
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/16
suffer. We shall later see how a page replacement policy can affect performance of a
system.
4.11.1 Page Replacement Policy
Towards understanding page replacement policies we shall consider a simple example of
a process P which gets an allocation of four pages to execute. Further, we assume that the
OS collects some information (depicted in Figure 4.10) about the use of these pages as
this process progresses in execution. Let us examine the information depicted in figure
4.10 in some detail to determine how this may help in evolving a page replacement
policy. Note that we have the following information available about P.
1. The time of arrival of each page. We assume that the process began at some time
with value of time unit 100. During its course of progression we now have pages
that have been loaded at times 112, 117 119, and 120.
2. The time of last usage. This indicates when a certain page was last used. This
entirely depends upon which part of the process P is being executed at any time.
Figure 4.10: Information on page usage policy.
3. The frequency of use. We have also maintained the frequency of use over some
fixed interval of time T in the immediate past. This clearly depends upon the
nature of control flow in process P.
As an example we may say that page located at 23 which was installed at time 119, was
last used at time unit 125 and over the time period T the process P made two references to
it. Based on the above pieces of information if we now assume that at time unit 135 the
process P experiences a page-fault, what should be done. Based on the choice of the
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/17
policy and the data collected for P, we shall be able to decide which page to swap out to
bring in a new page.
FIFO policy: This policy simply removes pages in the order they arrived in the main
memory. Using this policy we simply remove a page based on the time of its arrival in
the memory. Clearly, use of this policy would suggest that we swap page located at 14 as
it arrived in the memory earliest.
LRU policy: LRU expands to least recently used. This policy suggests that we re- move
a page whose last usage is farthest from current time. Note that the current time is 135
and the least recently used page is the page located at 23. It was used last at time unit 125
and every other page is more recently used. So, page 23 is the least recently used page
and so it should be swapped if LRU replacement policy is employed.
NFU policy: NFU expands to not frequently used. This policy suggests to use the
criterion of the count of usage of page over the interval T. Note that process P has not
made use of page located at 9. Other pages have a count of usage like 2, 3 or even 5
times. So the basic argument is that these pages may still be needed as compared to the
page at 9. So page 9 should be swapped.
Let us briefly discuss the merits of choices that one is offered. FIFO is a very simple
policy and it is relatively easy to implement. All it needs is the time of arrival. However,
in following such a policy we may end up replacing a page frame that is referred often
during the lifetime of a process. In other words, we should examine how useful a certain
page is before we decide to replace it. LRU and NFU policies are certainly better in that
regard but as is obvious we need to keep the information about the usage of the pages by
the process. In following the not frequently used (NFU) and least recently used (LRU)
page replacement policies, the OS needs to define recency. As we saw recency is defined
as a fixed time interval proceeding the current time. With a definition of recency, we can
implement the policy framework like least recently used (LRU). So one must choose a
proper interval of time. Depending upon the nature of application environment and the
work load a choice of duration of recency will give different throughput from the system.
Also, this means that the OS must keep a tab on the pages which are being used and how
often these are in use. It is often the case that the most recently used pages are likely to be
the ones used again. On the whole one can sense that the LRU policy should be
statistically better than FIFO.
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/18
A more advanced technique of page replacement policy may look-up the likely future
references to pages. Such a policy frame would require use of some form of predictive
techniques. In that case, one can prevent too many frequent replacements of pages which
prevents thrashing as discussed in the subsection. 4.11.2.
Let us for now briefly pay our attention to page references resulting in a page hit and a
page miss. When we find that a page frame reference is in the main memory then we
have a page hit and when page fault occurs we say we have a page miss. As is obvious
from the discussion, a poor choice of policy may result in lot of page misses. We should
be able to determine how it influences the throughput of a system. Let us assume that we
have a system with the following characteristics.
􀂾 Time to look-up page table: 10 time units.
􀂾 Time to look-up the information from a page frame (case of a page hit): 40 time
units.
􀂾 Time to retrieve a page from disk and load it and finally access the page frame
(case of a page miss): 190 time units.
Now let us consider the following two cases when we have 50% and 80% page hits. We
shall compute the average time to access.
􀂾 Case 1: With 50% page hits the average access time is ((10+40) * 0:5) + (10+190)
* 0:5 =125 time units.
􀂾 Case 2: With 80% page hits the average access time is (10+40) * 0:8) + (10+190)
* 0:2 = 80 time units.
Clearly, the case 2 is better. The OS designers attempt to offer a page replacement policy
which will try to minimize the page miss. Also, sometimes the system programmers have
to tune an OS to achieve a high efficacy in performance by ensuring that page miss cases
are within some tolerable limits. It is not unusual to be able to achieve over 90% page hits
when the application profile is very well known.
There is one other concern that may arise with regard to page replacement. It may be that
while a certain process is operative, some of the information may be often required.
These may be definitions globally defined in a program, or some terminal related IO
information in a monitoring program. If this kind of information is stored in certain pages
then these have to be kept at all times during the lifetime of the process. Clearly, this
requires that we have these pages identified. Some programming environments allow
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/19
directives like keep to specify such information to be available at all the time during the
lifetime of the process. In Windows there is a keep function that allows one to specify
which programs must be kept at all the time. The Windows environment essentially uses
the keep function to load TSR (terminate and stay resident) programs to be loaded in the
memory 1. Recall, earlier we made a reference to thrashing which arises from the
overheads generated from frequent page replacement. We shall next study that.
4.11.2 Thrashing
Suppose there is a process with several pages in its resident set. However, the page
replacement policy results in a situation such that two pages alternatively move in and out
of the resident set. Note that because pages are moved between main memory and disk,
this has an enormous overhead. This can adversely affect the throughput of a system. The
drop in the level of system throughput resulting from frequent page replacement is called
thrashing. Let us try to comprehend when and how it manifests. Statistically, on
introducing paging we can hope to enhance multi-programming as well as locality of
reference. The main consequence of this shall be enhanced processor utilization and
hence, better throughput. Note that the page size influences the number of pages and
hence it determines the number of resident sets we may support. With more programs in
main memory or more pages of a program we hope for better locality of reference. This is
seen to happen (at least initially) as more pages are available. This is because, we may
have more effective locality of reference as well as multi-programming. However, when
the page size becomes too small we may begin to witness more page-faults.
Incidentally, a virus writer may employ this to mount an attack. For instance, the keep
facility may be used to have a periodic display of some kind on the victim's screen. More
page-faults would result in more frequent disk IO. As disk IO happens more often the
throughput would drop. The point when this begins to happen, we say thrashing has
occurred. In other words, the basic advantage of higher throughput from a greater level of
utilization of processor and more effective multi-programming does not accrue any more.
When the advantage derived from locality of reference and multi-programming begins to
vanish, we are at the point when thrashing manifests. This is shown in Figure 4.11.
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/20
Figure 4.11: Thrashing on numerous page fault.
4.12 Paging: HW support
Recall we emphasized that we need HW within CPU to support paging. The CPU
generates a logical address which must get translated to a physical address. In Figure 4.12
we indicate the basic address generation and translation.
Let us trace the sequence of steps in the generation of address.
􀂾 The process generates a logical address. This address is interpreted in two parts.
􀂾 The first part of the logical address identifies the virtual page.
􀂾 The second part of the logical address gives the offset within this page.
􀂾 The first part is used as an input to the page table to find out the following:
* Is the page in the main memory?
* What is the page frame number for this virtual page?
􀂾 The page frame number is the first part of the physical memory address.
􀂾 The offset is the second part of the correct physical memory location.
Figure 4.12: Hardware support for paging.
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/21
If the page is not in the physical memory, a page-fault is generated. This is treated as a
trap. The trap then suspends the regular sequence of operations and fetches the required
page from the disk into main memory.
We next discuss a relatively simple extension of the basic paging scheme with hardware
support. This scheme results in considerable improvement in page frame access.
Figure 4.13: Paging with translation look-aside buffer.
4.12.1 The TLB scheme
The basic idea in the translation look-aside buffer access is quite simple. The scheme is
very effective in improving the performance of page frame access. The scheme employs a
cache buffer to keep copies of some of the page frames in a cache buffer. This buffer is
also interrogated for the presence of page frame copy. Note that a cache buffer is
implemented in a technology which is faster than the main memory technology. So, a
retrieval from the cache buffer is faster than that from the main memory. The hardware
signal which looks up the page table is also used to look up (with address translation) to
check if the cache buffer on a side has the desired page. This nature of look-up explains
why this scheme is called Translation Look-aside Buffer (TLB) scheme. The basic TLB
buffering scheme is shown in Figure 4.13. Note that the figure replicates the usual
hardware support for page table look-up. So, obviously the scheme cannot be worse than
the usual page table look-up schemes. However, since a cache buffer is additionally
maintained to keep some of the frequently accessed pages, one can expect to achieve an
improvement in the access time required for those pages which obtain a page hit for
presence in the buffer. Suppose we wish to access page frame p. The following three
possibilities may arise:
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/22
1. Cache presence: There is a copy of the page frame p. In this case it is procured
from the look-aside buffer which is the cache.
2. Page table presence: The cache does not have a copy of the page frame p, but
page table access results in a page hit. The page is accessed from the main
memory.
3. Not in page table: This is a case when the copy of the page frame is neither in the
cache buffer nor does it have an entry in the page table. Clearly, this is a case of
page-fault. It is handled exactly as the page-fault is normally handled.
Note that if a certain page frame copy is available in the cache then the cache look-up
takes precedence and the page frame is fetched from the cache instead of fetching it from
the main memory. This obviously saves time to access the page frame. In the case the
page hit occurs for a page not in cache then the scheme ensures its access from the main
memory. So it is at least as good as the standard paging scheme with a possibility of
improvement whenever a page frame copy is in cache buffer.
4.12.2 Some Additional Points
Since page frames can be loaded anywhere in the main memory, we can say that paging
mechanism supports dynamic relocation. Also, there are other schemes like multi-level
page support systems which support page tables at multiple levels of hierarchy. In
addition, there are methods to identify pages that may be shared amongst more than one
process. Clearly, such shareable pages involve additional considerations to maintain
consistency of data when multiple processes try to have read and write access. These are
usually areas of research and beyond the scope of this book.
4.13 Segmentation
Like paging, segmentation is also a scheme which supports virtual memory concept.
Segmentation can be best understood in the context of a program's storage requirements.
One view could be that each part like its code segment, its stack requirements (of data,
nested procedure calls), its different object modules, etc. has a contiguous space. This
space would then define a process's space requirement as an integrated whole (or
complete space). As a view, this is very uni-dimensional.
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/23
Figure 4.14: Segmentation scheme: A two dimensional view.
In using segmentation, one recognizes various segments such as the stack, object code,
data area etc. Each segment has requirements that vary over time. For instance, stacks
grow and shrink the memory requirements of object and data segments may change
during the lifetime of the process. This may depend on which functions have been called
and are currently active. It is, therefore, best to take a two-dimensional view of a
process's memory requirement. In this view, each of the process segments has an
opportunity to acquire a variable amount of space over time. This ensures that one area
does not run into the space of any other segment. The basic scheme is shown in Figure
4.14. The implementation of segmentation is similar to paging, except that we now have
segment table (in place of a page table) look-ups to identify addresses in each of the
segments. HW supports a table look-up for a segment and an offset within that segment.
We may now compare paging with segmentation.
􀂾 Paging offers the simplest mechanism to effect virtual addressing.
􀂾 While paging suffers from internal fragmentation, segmentation suffers from
external fragmentation.
􀂾 One of the advantages segmentation clearly offers is separate compilation of each
segment with a view to link up later. This has another advantage. A user may
develop a code segment and share it amongst many applications. He generates the
required links at the time of launching the application. However, note that this
also places burden on the programmer to manage linking. To that extent paging
offers greater transparency in usage.
􀂾 In paging, a process address space is linear. Hence, it is uni-dimensional. In a
segment based scheme each procedure and data segment has its own virtual space
mapping. Thus the segmentation assures a much greater degree of protection.
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/24
􀂾 In case a program's address space fluctuates considerably, paging may result in
frequent page faults. Segmentation does not suffer from such problems.
􀂾 Paging partitions a program and data space uniformly and is, therefore, much
simpler to manage. However, one cannot easily distinguish data space from
program space in paging. Segmentation partitions process space requirements
according to a logical division of the segments that make up the process.
Generally, this simplifies protection.
Clearly, a clever scheme with advantages of both would be: segmentation with paging. In
such a scheme each segment would have a descriptor with its pages identified. Such a
scheme is shown in Figure 4.15. Note that we have to now use three sets of offsets. First,
a segment offset helps to identify the set of pages. Next, within the corresponding page
table (for the segment), we need to identify the exact page table. This is done by using the
page table part of the virtual address. Once the exact page has been identified, the offset
is used to obtain main memory address reference. The final address resolution is exactly
as we saw in Section 4.9 where we first discussed paging.
Figure 4.15: Segmentation with paging.
In practice, there are segments for the code(s), data, and stack. Each segment carries the
rwe information as well. Usually the stack and data have read write permissions but no
Operating Systems/Memory management Lecture Notes
PCP Bhatt/IISc, Bangalore M4/V1/June 04/25
execute permissions. Code rarely has write permission but would have a read and execute
permission