Transcript CS 519 Operating Systems Theory Spring 1998

CMSC 412
Operating Systems
Fall 2002
Liviu Iftode
[email protected]
University of Maryland, CMSC 412, Fall 2002
1
Course overview
Goals





Understand how an operating system works as a
mediator between the computer architecture and
user programs
Learn how OS concepts are implemented in a real
operating system
Introduce to systems programming
Learn about performance evaluation
Learn about current trends in OS research
University of Maryland, CMSC 412, Fall 2002
2
OS Learning
project
OS Concepts
(lectures, textbooks)
OS Implementation
(source code, project doc)
recitations
Real OS
homeworks
(Unix/Linux textbooks)
OS Programming
(man pages)
University of Maryland, CMSC 412, Fall 2002
3
Course Timeline
Lecture
Concept A
Concept B
Tu
Tu
Th
Concept C
Th Tu
Apply A
Th
Apply B
Apply C
Lab
W
W
W
HW A Due
HW BC Due
Homework
W
W
Project AB Due
Project
W
University of Maryland, CMSC 412, Fall 2002
4
Suggested Approach





Read the assigned chapter from the textbook before
the lecture to understand the basic idea and become
familiar with the terminology
Attend the recitation
Start homeworks and project right away, systems
programming is not easy !
Ask questions during lecture, recitation.
Use the mailing list/newsgroup for discussions, do not
be afraid to answer a question posted by your
colleague even if you are not sure. This is a way to
validate your understanding of the material. Do not
forget, questions and discussions are not graded !
University of Maryland, CMSC 412, Fall 2002
5
Course Outline








Processes and process management
Threads and thread programming
Synchronization and Deadlock
Memory management and Virtual Memory
CPU Scheduling
File systems and I/O Management
Networking and Distributed Systems
Security
University of Maryland, CMSC 412, Fall 2002
6
Course requirements
Prerequisites


computer architecture
good programming skills (C!!, C++, Java)
Expected work




substantial readings (textbooks and papers)
challenging project extended over the entire
semester
homeworks (require programming)
midterm and final exams
University of Maryland, CMSC 412, Fall 2002
7
Work Evaluation




midterm exam 25%
homework 25%
project 25%
final exam 25%
University of Maryland, CMSC 412, Fall 2002
8
Homeworks
Goals



Deepen the understanding of OS concepts
Develop systems programming skills: virtual
memory, threads, synchronization, sockets
Learn to design, implement, debug and evaluate
the performance of an OS-bound program
Structure


4-5 homeworks
Both theoretical and C-programming problems
University of Maryland, CMSC 412, Fall 2002
9
Project
Goals

learn to design, implement and evaluate basic OS
mechanisms and policies
Structure



individual project
multiple phases
project report for each phase
University of Maryland, CMSC 412, Fall 2002
10
Textbooks



Stallings
Operating Systems. Internals and
Design Principles, 4th Edition,
Prentice-Hall, 2001.
Silberschatz, Galvin and Gagne, Operating System
Concepts, 6th Edition, John
Wiley & Sons, 2001.
Papers will be made available on the course
homepage
University of Maryland, CMSC 412, Fall 2002
11
Logistics






TAs:
Chunyuan Liao
Iulian Neamtiu (to be confirmed)
Course homepage:
http://www.cs.umd.edu/~iftode/cs412/cs412syllabus.htm
Preliminary schedule and course notes are already
available for the entire semester but they may be updated
before each class.
Homeworks every other week.
A project phase every other week.
A mailing list/newsgroup will be announced shortly.
University of Maryland, CMSC 412, Fall 2002
12
What is an operating system
application (user)
operating system
hardware


a software layer between the hardware and the
application programs/users which provides a virtual
machine interface: easy and safe
a resource manager that allows programs/users to
share the hardware resources: fair and efficient
University of Maryland, CMSC 412, Fall 2002
13
How does an OS work
application (user)
system calls
upcalls
commands
interrupts
hardware independent
OS
hardware dependent
hardware




receives requests from the application: system calls
satisfies the requests: may issue commands to hardware
handles hardware interrupts: may upcall the application
OS complexity: synchronous calls + asynchronous events
University of Maryland, CMSC 412, Fall 2002
14
Files
A file is a storage abstraction
application/user:
copy file1 file2
operating system: files, directories
hardware:


disk
naming, protection,
operations on files
operations on disk
blocks...
traditional approach: OS does disk block allocation and
caching (buffer cache) , disk operation scheduling and
replacement in the buffer cache
new approaches: application-controlled cache
replacement, log-based allocation (makes writes fast)
University of Maryland, CMSC 412, Fall 2002
15
Traditional file system
application:
read/write files
translate file to disk blocks
OS:
maintains
...buffer cache ...
controls disk accesses: read/write blocks
hardware:
University of Maryland, CMSC 412, Fall 2002
16
Mechanism vs Policy
application (user)
operating system: mechanism+policy
hardware



mechanism: data structures and operations that
implement the abstraction (e.g. the buffer cache)
policy: the procedure that guides the selection of a
certain course of action from among alternatives (e.g. the
replacement policy for the buffer cache)
traditional OS is rigid: mechanism together with policy
University of Maryland, CMSC 412, Fall 2002
17
Mechanism-policy split


traditional OS cannot provide the best policy in all cases
new OS approaches separate mechanisms from
policies:



OS provides the mechanism + some policy
applications contribute to the policy
flexibility+efficiency require new OS structures and/or
new OS interfaces
University of Maryland, CMSC 412, Fall 2002
18
Application-controlled caching
application:
read/write files
replacement policy
translate file to disk blocks
OS:
maintains
...buffer cache ...
controls disk accesses: read/write blocks
hardware:
University of Maryland, CMSC 412, Fall 2002
19
Processes
A process is a processor abstraction
user:
run application
operating system: processes
create, kill processes,
inter-process comm
context switch
hardware:


processor
traditional approach: OS switches processes on the
processor (process scheduling), provides inter-process
communication and handles exceptions
new approaches: application-controlled scheduling,
reservation-based scheduling, agile applications
University of Maryland, CMSC 412, Fall 2002
20
Traditional approach
active
processes
IPC
OS
processor



OS mediates inter-process communication (IPC)
OS schedules processes on the processor
application provides hints: priorities
University of Maryland, CMSC 412, Fall 2002
21
Hierarchical scheduling
processes
schedulers
OS
processor

OS schedules schedulers which schedule dependent
processes
University of Maryland, CMSC 412, Fall 2002
22
Virtual memory
Virtual memory is a memory abstraction
application:
address space
virtual addresses
operating system: virtual memory
physical addresses
hardware:


physical memory
traditional approach: OS provides a sufficiently large virtual
address space for each running application, does memory
allocation and replacement and may ensure protection
new approaches: external memory management, huge (64bit) address space, global memory
University of Maryland, CMSC 412, Fall 2002
23
VM: mechanism and policy
virtual address spaces
p1
p2
processes:
v-to-p memory mappings
physical memory:




processes can run being partially loaded in memory
illusion of more memory than physically available: swapping
processes can share physical memory if permitted
replacement policy can be exposed to the application
University of Maryland, CMSC 412, Fall 2002
24
Communication
Message passing is a communication abstraction
application:
sockets
naming, messages
operating system: TCP/IP protocols
network packets
hardware:


network interface
traditional approach: OS provides naming schemes, reliable
transport of messages, packet routing to destination
new approaches: zero-copy protocols, active messages,
memory-mapped communication
University of Maryland, CMSC 412, Fall 2002
25
Traditional OS structure

monolithic/layered systems

one/N layers all executed in “kernel-mode”

good performance but rigid
user
user
process
system calls
OS kernel
file
system
memory
system
hardware
University of Maryland, CMSC 412, Fall 2002
26
Micro-kernel OS
client
process
user mode
micro-kernel
file
server
memory
server
IPC
hardware





client-server model, IPC between clients and servers
the micro-kernel provides protected communication
OS functions implemented as user-level servers
flexible but efficiency is the problem
easy to extend for distributed systems
University of Maryland, CMSC 412, Fall 2002
27
Extensible OS kernel
process
process
my
memory
service
default
memory
service
user mode
extensible kernel


hardware
user processes can load customized OS services into
the kernel
good performance but protection and scalability
become problems
University of Maryland, CMSC 412, Fall 2002
28
Virtual Machines

old concept which is heavily revived today


the real hardware in “cloned” in several
identical virtual machines
OS functionality built on top of the virtual
machine
user
OS on virtual machine
allocate resource
exokernel
hardware
University of Maryland, CMSC 412, Fall 2002
29
Computer System





Processor: performs data processing
Main memory: stores both data and programs,
typically volatile
Disks: secondary memory devices which provide
persistent storage
Network interfaces: inter-machine communication
Buses: intra-machine communication


memory bus (processor-memory)
I/O bus (disks, network interfaces, other I/O
devices, memory-bus)
University of Maryland, CMSC 412, Fall 2002
30
Basic computer structure
CPU
Memory
memory bus
I/O bus
disk
Net interface
University of Maryland, CMSC 412, Fall 2002
31
Memory caches






motivated by the mismatch between processor and
memory speed
closer to the processor than the main memory
smaller and faster than the main memory
act as “attraction memory”: contains the value of
main memory locations which were recently
accessed (temporal locality)
transfer between caches and main memory is
performed in units called cache blocks/lines
caches contain also the value of memory locations
which are close to locations which were recently
accessed (spatial locality)
University of Maryland, CMSC 412, Fall 2002
32
Cache design issues
cpu
word transfer
cache
block transfer
main
memory




cache size and cache block size
mapping: physical/virtual caches, associativity
replacement algorithm: LRU
write policy: write through/write back
University of Maryland, CMSC 412, Fall 2002
33
Memory Hierarchy
cpu
word transfer
cache
block transfer
main memory
page transfer





disks
decrease cost per bit
decrease frequency of
access
increase capacity
increase access time
increase size of
transfer unit
University of Maryland, CMSC 412, Fall 2002
34
Data transfer on the bus
CPU
cache
Memory
memory bus
I/O bus
disk




Net interface
cache-memory: cache misses, write-through/write-back
memory-disk: swapping, paging, file accesses
memory-Network Interface : packet send/receive
I/O devices to the processor: interrupts
University of Maryland, CMSC 412, Fall 2002
35
Direct Memory Access (DMA)

bulk data transfer between memory and an I/O device
(disk, network interface) initiated by the processor







address of the I/O device
starting location in memory
number of bytes
direction of transfer (read/write from/to
memory)
processor interrupted when the operation completes
bus arbitration between cache-memory and DMA
transfers
memory cache must be consistent with DMA
University of Maryland, CMSC 412, Fall 2002
36
Multiprocessors
CPU
cache
CPU
cache
Memory
memory bus
I/O bus
disk





Net interface
simple scheme: more than one processor on the same bus
memory is shared among processors-- cache consistency
bus contention increases -- does not scale
alternative (non-bus) system interconnect -- expensive
single-image operating systems
University of Maryland, CMSC 412, Fall 2002
37
Multicomputers
CPU
cache
CPU
cache
Memory
Memory
memory bus
memory bus
I/O bus
I/O bus
network
disk




Net interface
Net interface
disk
network of computers: “share-nothing” -- cheap
communication through message-passing: difficult to program
challenge: build efficient shared memory abstraction in
software
each system runs its own operating system
University of Maryland, CMSC 412, Fall 2002
38