421/521 CSE - Operating Systems
Fall 2011
Lecture - II
OS Structures
Tevfik Ko!ar
University at Buffalo
September 1st, 2011
1
Roadmap
OS Design and Implementation
Different Design Approaches
Major OS Components
!
!
!
!
Processes
Memory management
CPU Scheduling
I/O Management
2
OS Design Approaches
3
Operating System Design and Implementation
Start by defining goals and specifications
Affected by choice of hardware, type of system
Batch, time shared, single user, multi user, distributed
User goals and System goals
User goals operating system should be convenient to use, easy
to learn, reliable, safe, and fast
System goals operating system should be easy to design,
implement, and maintain, as well as flexible, reliable, errorfree, and efficient
No unique solution for defining the requirements of an
OS
! Large variety of solutions
! Large variety of OS
4
Operating System Design and Implementation (Cont.)
Important principle: to separate policies and
mechanisms
Policy: What will be done?
Mechanism: How to do something?
Eg. to ensure CPU protection
Use Timer construct (mechanism)
How long to set the timer (policy)
The separation of policy from mechanism allows
maximum flexibility if policy decisions are to be
changed later
5
OS Design Approaches
Simple Structure (Monolithic)
Layered Approach
Microkernels
Modules
6
Simple Structure
Monolithic
No well defined structure
Start as small, simple, limited systems, and then grow
No Layers, not divided into modules
7
Simple Structure
" Example: MS-DOS
#
initially written to provide
the most functionality in
the least space
#
started small and grew
beyond its original scope
#
levels not well separated:
programs could access I/O
devices directly
#
excuse: the hardware of
that time was limited (no
dual user/kernel mode)
Silberschatz, A., Galvin, P. B. and Gagne. G. (2003)
Operating Systems Concepts with Java (6th Edition).
MS-DOS pseudolayer structure
8
Layered Approach
" Monolithic operating systems
# no one had experience in building truly large software systems
# the problems caused by mutual dependence and interaction
were grossly underestimated
# such lack of structure became unsustainable as O/S grew
" Enter hierarchical layers and information abstraction
# each layer is implemented exclusively using operations
provided by lower layers
# it does not need to know how they are implemented
# hence, lower layers hide the existence of certain data
structures, private operations and hardware from upper layers
9
"
Simple Layered Approach
The original UNIX
#
enormous amount of
functionality crammed
into the kernel everything below
system call interface
#
The Big Mess: a
collection of
procedures that can
call any of the other
procedures whenever
they need to
#
no encapsulation, total
visibility across the
system
#
very minimal layering
made of thick,
monolithic layers
Silberschatz, A., Galvin, P. B. and Gagne. G. (2003)
Operating Systems Concepts with Java (6th Edition).
UNIX system structure
10
Layered Approach
The operating system is divided
into a number of layers (levels),
each built on top of lower
layers.
The bottom layer (layer 0), is the
hardware;
The highest (layer N) is the user
interface.
With modularity, layers are
selected such that each uses
functions (operations) and
services of only lower-level
layers
GLUnix, Multics, VAX/VMS
11
Layered Approach
"
Layers can be debugged and replaced independently
without bothering the other layers above and below
#
# famous example of
strictly layered
architecture: the
TCP/IP networking
stack
N+1
offers services
N
uses services
N1
12
Layered Approach
Theoretical model of operating system design hierarchy
shell
O/S
hardware
Stallings, W. (2004) Operating Systems:
Internals and Design Principles (5th Edition).
13
Layered Approach
" Major difficulty with layering
# . . . appropriately defining the various layers!
# layering is only possible if all function dependencies can be
sorted out into a Directed Acyclic Graph (DAG)
# however there might be conflicts in the form of circular
dependencies (cycles)
Circular dependency on top of a DAG
14
Layered Approach
" Circular dependencies in an O/S organization
# example: disk driver routines vs. CPU scheduler routines
$ the device driver for the backing store (disk space used by
virtual memory) may need to wait for I/O, thus invoke the
CPU-scheduling layer
$ the CPU scheduler may need the backing store driver for
swapping in and out parts of the table of active processes
" Other difficulty: efficiency
# the more layers, the more indirections from function to function
and the bigger the overhead in function calls
# backlash against strict layering: return to fewer layers with
more functionality
15
Microkernel System Structure
" The microkernel approach
# a microkernel is a reduced operating system core that contains
only essential O/S functions
# the idea is to minimize the kernel by moving up as much
functionality as possible from the kernel into user space
# many services traditionally included in the O/S are now external
subsystems running as user processes
$ device drivers
$ file systems
$ virtual memory manager
$ windowing system
$ security services, etc.
Examples: QNX, Tru64 UNIX, Mach (CMU), Windows NT
16
Layered OS vs Microkernel
17
Microkernel System Structure
" Benefits of the microkernel approach
#
#
#
extensibility it is easier to extend a microkernel-based O/S as new
services are added in user space, not in the kernel
portability it is easier to port to a new CPU, as changes are needed only
in the microkernel, not in the other services
reliability & security much less code is running in kernel mode; failures
in user-space services dont affect kernel space
" Detriments of the microkernel approach
#
#
again, performance overhead due to communication from user space to
kernel space
not always realistic: some functions (I/O) must remain in kernel space,
forcing a separation between policy and mechanism
18
Modular Approach
" The modular approach
# many modern operating systems implement kernel modules
# this is similar to the object-oriented approach:
$ each core component is separate
$ each talks to the others over known interfaces
$ each is loadable as needed within the kernel
# overall, modules are similar to layers but with more flexibility
# modules are also similar to the microkernel approach, except
they are inside the kernel and dont need message passing
19
Modular Approach
"
Modules are used in Solaris, Linux and Mac OS X
Silberschatz, A., Galvin, P. B. and Gagne. G. (2003)
Operating Systems Concepts with Java (6th Edition).
The Solaris loadable modules
20
Mac OS X Structure - Hybrid
BSD: provides support for command line interface, networking, POSIX file
system, API and threads
Mach: memory management, RPC, IPC, message passing
21
Major OS Components
22
Major OS Components
%
%
%
%
Processes
Memory management
CPU Scheduling
I/O Management
23
Processes
" A process is the activity of executing a program
Pasta for six
boil 1 quart salty
water
CPU
thread of execution
stir in the pasta
cook on medium
until al dente
serve
Program
input data
Process
24
Processes
"
It can be interrupted to let the CPU execute a higher-priority
process
Pasta for six
First aid
CPU (changes hat to doctor)
boil 1 quart salty kit
Get the first aid
water
Check pulse
thread of execution
stir in Clean wound with
the pasta
alcohol
input data
cook on medium
Apply band
untilal denteaid
serve
Process
Program
25
Processes
"
. . . and then resumed exactly where the CPU left off
hmm... now
where was
I?
Pasta for six
boil 1 quart salty
water thread of execution
CPU (back to
chef)
stir in the pasta
cook on medium
until al dente
serve
Program
input data
Process
26
Processes
job 3
job 1
...
job 1
job 4
job 3
job 1
job 2
...
job 1
Multitasking gives the illusion of parallel processing
(independent virtual program counters) on one CPU
job 2
"
...
job 2
job 1
process 1
process 2
job 3
job 2
...
job 1
(a) Multitasking from the CPUs viewpoint
job 3
job 4
process 3
process 4
(b) Multitasking from the processes viewpoint = 4 virtual program counters
Pseudoparallelism in multitasking
27
Processes
Timesharing is logical extension in which CPU switches
jobs so frequently that users can interact with each job
while it is running, creating interactive computing
Response time should be < 1 second
Each user has at least one program loaded in memory and
executing ! process
28
Processes
" Operating System Responsibilities:
The O/S is responsible for managing processes
#
the O/S creates & deletes processes
#
the O/S suspends & resumes processes
#
the O/S schedules processes
#
the O/S provides mechanisms for process synchronization
#
the O/S provides mechanisms for interprocess
communication
#
the O/S provides mechanisms for deadlock handling
29
Memory Management
" Operating System Responsibilities:
The O/S is responsible for an efficient and orderly
control of storage allocation
#
ensures process isolation: it keeps track of which parts of
memory are currently being used and by whom
#
allocates and deallocates memory space as needed: it
decides which processes to load or swap out
#
regulates how different processes and users can
sometimes share the same portions of memory
#
transfers data between main memory and disk and ensures
long-term storage
30
Memory Management
" Main memory
"
#
large array of
words or bytes,
each with its own
address
#
repository of
quickly accessible
data shared by
the CPU and I/O
devices
#
volatile storage
that loses its
contents in case
of system failure
The storage hierarchy
31
Performance of Various Levels of Storage
Movement between levels of storage hierarchy can be
explicit or implicit
32
Caching
Important principle, performed at many levels in a
computer (in hardware, operating system, software)
Information in use copied from slower to faster storage
temporarily
Faster storage (cache) checked first to determine if
information is there
If it is, information used directly from the cache (fast)
If not, data copied to cache and used there
If cache is smaller than storage being cached
Cache management - important design problem
Cache size and replacement policy
33
Migration of Integer A from Disk to Register
Multitasking environments must be careful to use most
recent value, not matter where it is stored in the
storage hierarchy
Multiprocessor environment must provide cache
coherency in hardware such that all CPUs have the
most recent value in their cache
Distributed environment situation even more complex
Several copies of a datum can exist
34
CPU Scheduling
" Operating System Responsibilities:
The O/S is responsible for efficiently using the CPU
and providing the user with short response times
#
decides which available processes in memory are to be
executed by the processor
#
decides what process is executed when and for how long,
also reacting to external events such as I/O interrupts
#
relies on a scheduling algorithm that attempts to optimize
CPU utilization, throughput, latency, and/or response time,
depending on the system requirements
35
OS Scheduling
the decision to add a program to the pool of
processes to be executed (job scheduling)
" Medium-term scheduling
#
the decision to add to the number of processes that
are partially or fully in main memory (swapping)
" Short-term scheduling = CPU scheduling
#
the decision as to which available processes in
memory are to be executed by the processor
(dispatching)
" I/O scheduling
#
the decision to handle a processs pending I/O
request
frequency of intervention
#
fine- to coarse-grain level
" Long-term scheduling
I/O Management
" Operating System Responsibilities:
The O/S is responsible for controlling access to all
the I/O devices
#
hides the peculiarities of specific hardware devices from
the user
#
issues the low-level commands to the devices, catches
interrupts and handles errors
#
relies on software modules called device drivers
#
provides a device-independent API to the user programs,
which includes buffering
37
I/O Management
Device-independent software
Device-independent software
Tanenbaum, A. S. (2001)
Modern Operating Systems (2nd Edition).
Layers of the I/O subsystem
38
Two I/O Methods
After I/O starts, control returns to user program only
upon I/O completion & synchronous
Wait instruction idles the CPU until the next interrupt
Wait loop (contention for memory access).
At most one I/O request is outstanding at a time, no
simultaneous I/O processing.
After I/O starts, control returns to user program
without waiting for I/O completion &asynchronous
System call request to the operating system to allow user
to wait for I/O completion.
Device-status table contains entry for each I/O device
39
Two I/O Methods
Synchronous
Asynchronous
40
Summary
OS Design Approaches
Questions?
Monolithic Systems,
Layered Approach, Microkernels, Modules
Major OS Components
!
!
!
!
Processes
Memory management
CPU Scheduling
I/O Management
Reading Assignment: Chapter 2 from Silberschatz.
41
Acknowledgements
Operating Systems Concepts book and supplementary
material by A. Silberschatz, P. Galvin and G. Gagne
Operating Systems: Internals and Design Principles
book and supplementary material by W. Stallings
Modern Operating Systems book and supplementary
material by A. Tanenbaum
R. Doursat and M. Yuksel from UNR
42