P Part I Overview t I O i C Chapter 2: Operating System - - PowerPoint PPT Presentation
P Part I Overview t I O i C Chapter 2: Operating System Structures 2 O i S S 1 Fall 2010 Operating System Services Operating System Services User Interface ( e.g ., command processor shell and GUI Graphical User
1
Fall 2010
User Interface (e.g., command processor –shell and GUI – Graphical User Interface) and GUI Graphical User Interface) Program execution I/O operations I/O operations File-system manipulation Communications Communications Error detection Resource allocation Resource allocation Accounting Protection
2
Protection
System calls provide an interface to the services d il bl b i made available by an operating system. Type of system calls:
Process control (e.g., create and destroy processes) File management (e.g., open and close files) Device management (e.g., read and write operations) Information maintenance (e.g., get time or date) Communication (e.g., send and receive messages)
3
register X syscall 10 X syscall 10 service routine
syscall parameters
Load addr. X Load addr. X syscall 10
4
A system call generates kernel A system call generates an interrupt, actually a trap
interrupt h dl
syscall i
kernel trap. The executing program i d d
handler
services d it h
is suspended. Control is then
mode switch
transferred to the OS. Program continues
syscall 10
g when the system call service completes.
y
5
p
File management (e.g., create, delete files) Status management (e.g., date/time, available memory and disk space) File modification (e.g., editors) Programming language support (e.g., assemblers, compilers, interpreters) Program loading and execution (e.g., loaders, linkage editors) Communications (e.g., connections between processes, sending/receiving e-mails)
6
Design Goals Mechanisms and Polices Implementation Implementation
7
Design Goals
Goals and specifications are affected by hardware and the type of the system (e.g., b t h ti h i l ti t ) batch, time-sharing, real-time, etc) Requirements: user goals and system goals.
User goals: easy of use, reliable, safe, fast System goals: easy to design, implement and maintain and flexible reliable efficient etc maintain, and flexible, reliable, efficient, etc.
There is no unique way to achieve all goals, and some compromises must be taken
8
and some compromises must be taken.
Mechanisms and Policies: 1/2 Mechanisms and Policies: 1/2
Mechanisms determine how to do something. Policies determine what will be done. Policies and mechanisms are usually separated: the separation of mechanism and policy principle.
9
Mechanisms and Policies: 2/2
The separation of mechanism and policy: It is for efficiency purpose. Policies are likely to change over time. If l h i i il bl h i If a more general mechanism is available, the impact
Microkernel-based systems implement a basic set of Microkernel-based systems implement a basic set of building blocks, which are almost policy free. Thus, advanced policies and mechanisms can be added to d k l d l user-created kernel modules.
10
Implementation
The assembly vs. high-level language issue
Advantages of high-level languages: easy to use and i t i d il bl i t l maintain, and more available programming tools Disadvantages of high-level languages: slower and more space more space
Many systems were written in both with a small percentage (e.g., 10%) of assembly language for percentage (e.g., 10%) of assembly language for the most speed demanding low level portion. Better data structures and algorithms are more
11
Better data structures and algorithms are more important than tweaking assembly instructions.
Simple Structure Layered Approach Layered Approach Microkernels Modules
12
Simple structure systems do not have well-defined
MS-DOS structure
structures Unix only had a limited application programs structure: kernel and system programs resident sys program Everything between the system call interface and physical hardware is the
ROM BIOS device driver MS-DOS device drivers
physical hardware is the kernel.
ROM BIOS device driver
13
The operating system is broken up into a number of layers (or levels) each on top of number of layers (or levels), each on top of lower layers. Each layer is an implementation of an abstract y p
Th b tt l (l 0) i th h d d The bottom layer (layer 0) is the hardware, and the highest one is the user interface. The main advantage of layered approach is The main advantage of layered approach is modularity.
14
The lowest layer is process The lowest layer is process management. Each layer only uses the Users Each layer only uses the
layers and does not have to file system
y know their implementation. Each layer hides the communication I/O & Device
sor mo
y existence of certain data structures, operations and I/O & Device virtual mem
upervis
hardware from higher-level
h d process
su
15
hardware
It is difficult to organize the system in It is difficult to organize the system in layers, because a layer can use only layers below it Example: virtual layers below it. Example: virtual memory (lower layer) uses disk I/O (upper layer) (upper layer). Layered implementations tend to be less efficient than other types. Example: there may be too many calls going down the layers: user to I/O layer to memory layer to process scheduling layer.
16
The layered approach
layer n: user interface
The layered approach may also represented as a set of concentric rings
n
a set of concentric rings. The first OS based on th l d h
1
the layered approach was THE, developed by E Dijk t
17
layer 0: hardware
Only absolutely essential core OS functions should be in the kernel. Less essential services and applications are built on the kernel and run in user are built on the kernel and run in user mode. Many functions that were in a traditional Many functions that were in a traditional OS become external subsystems that interact ith the kernel and ith each interact with the kernel and with each
18
The main function The main function
to provide
pro virtu f dev clie
to provide communication facility between the
cess ser ual mem file serv vice dri ent pro
facility between the client programs and various services.
rver mory ver ivers
various services. Communication is provided by provided by message passing. hardware
microkernel
19
hardware
Uniform interfaces: message passing Uniform interfaces: message passing Extensibility: adding new services is easy i i i i i i Flexibility: existing services can be taken out easily to produce a smaller and more efficient implementation implementation Portability: all or at least most of the processor specific code is in the small kernel specific code is in the small kernel. Reliability: A small and modular designed kernel can be tested easily kernel can be tested easily Distributed system support: client and service processes can run on networked systems
20
processes can run on networked systems.
B t i k l d h bl But, microkernels do have a problem:
As the number of system functions increases,
Most microkernel systems took a hybrid
c o e e sys e s oo yb d approach, a combination of microkernel and something else (e.g., layered). g ( g , y )
21
Microkernel vs. Layered Approach Microkernel vs. Layered Approach
Users
pro virt de cli
file system
ual me file ser evice dr ient pro
communication I/O & Device
sor mo erver mory rver rivers
I/O & Device virtual mem
upervis
h d process
su
hardware
microkernel
22
hardware hardware
Th OO t h l The OO technology may be used to create a modular
scheduling
create a modular kernel. The kernel has a set
scheduling device/bus drivers file system
and dynamically
core Solaris kernel
loadable
links in additional services either during boot time or
modules syscalls
during boot time or during run time.
STREAMS modules executable format
23
The module approach looks like the layered h h d l h i d fi i i approach as each module has a precise definition; however, the module approach is more flexible in th t d l ll th d l that any module can call any other module. The module approach is also like the microkernel approach because the core module only includes the core functions. However, the module approach is more efficient because no message passing is required.
24
A virtual machine, VM, is a software between the kernel and hardware between the kernel and hardware. A VM provides all functionalities of a CPU with software simulation. A user has the illusion that s/he has a real A user has the illusion that s/he has a real processor that can run a kernel.
25
processes processes processes processes p kernel kernel kernel kernel kernel kernel kernel VM1 VM2 VM3 hardware e e hardware hardware VM implementation
26
S lf Vi t li d VM th VM i id ti l t th Self-Virtualized VM: the VM is identical to the
VM ( ti VM d Li t VMware (creating a VM under Linux to run Windows). Non-Self-Virtualized VM: the VM is not identical to the hardware. Example: Java Virtual Machine JVM and SoftWindow. It can be proved that all third-generation CPUs p g can be virtualized.
27
VM’s are difficult to implement because they VM s are difficult to implement because they must duplicate all hardware functions. Benefits: Benefits: VM provides a robust level of security VM permits system development to be done VM permits system development to be done without disrupting normal system operation. VM allows multiple operating systems to run VM allows multiple operating systems to run
VM can make system transition/upgrade VM can make system transition/upgrade much easier
28
VMware Java VM
29
When the system is powered on a small When the system is powered on, a small program, usually stored in ROM (i.e., read only memory) is executed This initial program is memory), is executed. This initial program is usually referred to as a bootstrap program or an Initial Program Loader (IPL). g ( ) The bootstrap program reads in and initializes the operating system. The execution is then p g y transferred to the OS
30
31