Operating Systems and Systems Programming Another view of change Professor Anthony D Joseph Spring 2004 Lecture 2 Operating System Structures 2 0 Administrivia 2 0 Main points More Many people per computer One person per computer Info Appliances PC Network Distribution WS Server History of operating systems Time Sharing Overview of Course Topics Overview of Main OS structures Batch Less Many computers per person RJE Time 2 1 History of Operating Systems Change What does this mean Techniques have to vary over time adapt to changing tradeoffs One view of change Typical academic computer in 1981 and 2004 2 1 1 History Phase 1 hardware expensive humans cheap 1948 1970 1981 2004 Factor CPU MHz cycles inst 10 3 4 3200 0 25 0 5 320 1 5 12 DRAM capacity 128KB 2 GB 16 000 Disk capacity 10 MB 500 GB 50 000 net bandwidth 9600 b s 1 Gb s 110 000 address bits 16 32 2 users mach 10s 1 0 1 Price 25000 5500 0 22 When computers cost millions of s optimize for more efficient use of the hardware Lack of interaction between user and computer User at console one user at a time Batch monitor load program run print Computers are expensive so optimize to make better use of the hardware When the user at the console is thinking he s not using the computer So instead feed the computer jobs in batches and make users wait for their results Makes better use of hardware but a lot worse for users LEAVE SPACE ON BOARD No protection what if batch program had a bug and wrote over batch monitor CS 162 Spring 2004 Lecture 2 1 1 Data channels interrupts overlap of I O and computation DMA Direct Memory Access for I O devices OS requests I O goes back to computing gets interrupt when I O device has finished CS 162 Spring 2004 Lecture 2 2 2 Multiprogramming several programs run at the same time users share system 2 1 3 History Phase 3 hardware very cheap humans very expensive 1981 1990 Multiprogramming benefits 1 Small jobs not delayed by large jobs Computer costs 1K Programmer costs 100K year If you can make someone 1 more efficient by giving them a computer it s worth it 2 More overlap between I O and CPU Multiprogramming requires memory protection to keep bugs in one program from crashing the system or corrupting other programs Bad news OS must manage all these interactions between programs Each step seems logical but at some point fall off cliff just gets too complicated Multics announced in 1963 ran in 1969 OS 360 released with 1000 bugs OS s finally begin to be an important science How do we deal with this Personal computing Computers are cheap so give everyone a PC Initially PCs had limited hardware resources so OS became a subroutine library for running one application at a time again MSDOS But since then PCs have become powerful computers and the OS has regained all the complexity of a big OS memory protection multiprogramming etc NT OS 2 complexity UNIX based on Multics but simplified so they could get it to work 2 1 2 Key challenge is to use computers to make more efficient use of people s time History Phase 2 hardware cheap humans expensive 1970 1985 Question As hardware gets cheaper and cheaper does the need for some or all OS functions go away 2 1 4 History Phase 4 Distributed systems 1989 Computers available for tens of thousands of dollars instead of millions Can Networking allow different machines to share resources easily printers file afford to start using them for more kinds of jobs Interactive timesharing Use cheap terminals 1 2K to let multiple users interact with the system at the same time Sacrifice CPU time to get better response time for users Users start doing things like program development and debugging test editing and e mail online servers web servers Client Server model 2 1 5 History Phase 5 Mobile systems 1995 Mobile devices laptops Personal Digital Assistants etc small portable and inexpensive but have limited capabilities memory CPU power etc If they have wireless support they can leverage resources of distributed systems Problem thrashing performance falls off a cliff as you add users Today s lightweight laptops are nearly as powerful as desktops CS 162 Spring 2004 Lecture 2 3 3 CS 162 Spring 2004 Lecture 2 4 4 2 2 High end lightweight laptop in 2004 2 1 6 Laptop 2004 CPU MHz cycles inst 1600 0 25 0 5 DRAM capacity 2GB Disk capacity 80GB net BW wired 1 Gb s wireless LAN wide area 54 Mb s 2 Mb s address bits 32 users mach 1 Price 3700 Overview of Course Topics The course will be divided into three main topics with exams corresponding roughly to these 1 Concurrency processes and threads 2 Memory management File Systems and I O 3 Distributed Systems and Networking 2 3 Operating System Structures What follows is a brief tour of the overall structure of OS s There are several ways to divide things up and each yields a different kind of understanding about an OS 2 3 1 System Components What are the pieces of the OS Process management Main memory management I O system management File management Networking UI History of Operating Systems Summary Point of change isn t look how stupid batch processing is It was right for the tradeoffs of the time but not anymore Point is have to change with changing technology 2 3 2 Operating System Services what things does the OS do Services that more or less map onto components Program execution I O operations File system manipulation Communications Situation today is much like it was in the late 60 s OS s today are enormous complex things small OS 100K lines big OS 10M lines Cross cutting capabilities Error detection recovery Resource allocation Accounting Protection 100 1000 people years NT under development from early 90 s to late 90 s similar to mainframe OS development cycles still doesn t work very well Jury is still out for Windows 2000 and Windows XP Windows Longhorn delayed until 2006 2 3 3 System Calls what s the API See Chap 3 of the book for an example list Key aspect of this course understand OS s so we can simplify them CS 162 Spring 2004 Lecture 2 5 5 CS 162 Spring 2004 Lecture 2 6 6 2 3 4 System Structure what s the implementation structure Simple structure Unstructured e g DOS Kernel user e g original Unix Layered approach Idea each layer uses only the routines of the layers below it Advantage modularity easier debugging and maintenance One of the most important layerings machine dependent versus portable layer Micro kernel plus server processes Small kernel plus user level servers Even more modular and debuggable and maintainable Very
View Full Document
Unlocking...