What is an operating system?

Size: px

Start display at page:

Download "What is an operating system?"

Jasmin Alexandra Burns
5 years ago
Views:

1 What is an operating system? slide 1 "Peterson and Silbershatz" "We can view anoperating system as a resource allocator. Acomputer system has many resources (hardware and software) that may be required to solve a problem: CPU time, memory space, file storage space, I/O devices, and so on. The operating system acts as the manager ofthese resources and allocates them to specific programs and users asnecessary for their tasks. Since there may be many, possibly conflicting requests for resources, the operating system must decide which requests are allocated resources to operate the computer system efficiently and fairly."

2 Windows NT slide 2 many elements influenced the design of Windows NT it provides the same type of GUI found in the earlier Windows products including the use of windows pull down menus point and click interaction

3 NT components slide 3 internal interaction is inspired by the Mach operating system, which is in turn based on UNIX highly modular can execute on a variety of hardware platforms supports applications written for a variety of other operating systems as with virtually all operating systems, NT separates application oriented software from operating system software. (Kernel mode and user mode)

4 Design goals of NT slide 4 like any commercial operating system, Windows NT is the result of a complex interaction between idealized goals and market-driven realities Compatibility it should support a wide range of existing software and legacy hardware. (DOS, Windows 3.1 compatibility) Robustness it should resist attacks of naive or malicious users, individual programs should be isolated from each other Portability it should run on a wide variety of current and future hardware platforms

6 Design goals of NT slide 6 Extendibility it should be possible to add new features and support new I/O devices Performance it should give reasonable performance on commonly available hardware. Take advantage of multiprocessing hardware

7 Major building blocks of NT slide 7 Overall architecture of the NT kernel mode components

8 Hardware abstraction layer slide 8 HAL is a thin layer of software that presents the rest of the system with an abstract model of any hardware that is not part of the CPU HAL exposes a well defined set of functions that manage such items as: Off chip caches Timers I/O buses Device registers Interrupt controllers DMA controllers various system components use these HAL functions to interact with CPU hardware HAL routines make the kernel and device drivers binary compatible across platforms with the same CPU architecture

10 Kernel slide 10 consists of the most fundamental components of the operating system manages scheduling and context switching exception and interrupt handling multiprocessor synchronization Executive the most complex kernel mode component in Windows NT implement many ofthe basic functions normally associated with an operating system (Process manager, virtual memory manager, I/O manager, system call interface) the executive uses the HAL to interact with any CPU hardware

11 Layered device drivers in NT slide 11 the term driver normally refers to a piece of code that manages some peripheral or device NT takes a more flexible approach allows several driver layers to exist between an application program and hardware Device drivers manage actual data transfer and control operations for a specific type of physical device includes starting stopping I/O, handling interrupts and handling errors Intermediate drivers

12 NT allows you to place any number of layers on top of a physical device driver intermediate layers provide a way of extending the capabilities of the I/O system without having to modify the drivers below them

13 Example: Fault tolerant disk driver slide 13 is implemented as a layer that sits between the file system and drivers for any physical disks. intermediate drivers separate hardware specific operations from more general management issues here the intermediate driver isreferred to as a class driver and the hardware driver iscalled a port driver this structure makes it easier to target a wider range of hardware since only the port code needs to be rewritten

14 File system drivers (FSDs) slide 14 the following FSDs are supplied by Microsoft: Type Description FAT NTFS HPFS CSFS MSFS NPFS RDR Windows 95 extended MSDOS file system Windows NT high reliability file system OS/2 high performance file system ISO 9660 CDROM file system Mailshot file system Named pipe file system LAN Manager redirector unfortunately, you cannot develop file system drivers using the standard NT device driver kit also note the absence of NFS.

16 Device driver slide 16

17 UNIX vs NT slide 17 NT available on other architectures Alpha, Mips, Power PC UNIX also available on other architectures, all of the above and more NT will run all Windows 3.1 software and MSDOS software UNIX may or may not depending on version, vendor generally unlikely

18 UNIX vs NT slide 18 both NT and UNIX (Solaris, Sequent) will operate with multiple processors both OSs have TCP/IP and associated utilities in different forms NT doesn t have NFS though (which considered slow, but open!) nevertheless much research has been done to improve NFS performance both OSs support filesystems with long filenames NT will easily integrate with Novell, MACOS UNIX needs extra packages (eg samba) which has to be configured by an expert

19 UNIX vs NT slide 19 UNIX is still more flexible in resource sharing share any peripheral across the network all peripherals are as files, files can be on the network etc Windows NT NTFS is faster than NFS UNIX buffer cache is superior to NT thus it is not clear which OS will be faster! measure it! TCP/IP code is more mature than NT, witness problems in J111

20 UNIX vs NT slide 20 NT is much easier to manage NT is much easier to install NT is more tolerant of different hardware devices due to the HAL

21 Network (Unix) slide 21 UNIX started life as a research project within AT&T in 1970 since that time it has grown in popularity and functionality there are some excellent books available which describe UNIX Berny Goodheart & James Cox, "The magic garden explained", Prentice-Hall. Smoot Carl-Mitchell & John S. Quarterman, "Practical Internetworking with TCP/IP and UNIX", Addison Wesley. W.Richard Stevens, "UNIX Network Programming", Prentice-Hall.

22 UNIX networking history slide 22 the first networking functionality added to UNIX wasuucp (UNIX to UNIX copy) in 1976 provided a mechanism to transfer files from different UNIX machines which typically only had modem connectivity implemented was a batch-oriented system that was typically used between systems remote printing and were added in the Berkeley release of the Version 7 UNIX for the PDP-11 (called 2.x BSD) (Berkeley Standard Distribution) TCP/IP was released in 1981 in the 4.1 BSD UNIX

23 Unix history slide 23 TCP/IP opened the flood gates for networked applications NFS - network file system remote job execution remote login, telnet, ftp, R tools talk remote procedure calls and XDR news client server computing

24 UNIX networking history slide 24 the Berkeley releases of UNIX 4.1 BSD and 4.2 BSD provided the first UNIX systems with widespread network support ( ) many Computer Science departments started to buy more than one VAXwith Ethernet cards and so created UNIX LANs April BSD was released June 1988 a revision of the 4.3 BSD referred to as the Tahoe distribution was meant to be an interim release for testing by experienced users importantly it contained source code which was not licensed by AT&T so you could legally use the source code for your own products reference software was available for TCP/IP & sockets

25 UNIX network services slide 25 rexd remote job execution daemon allows a user to say: on floppsie gcc -g largeprogram.c theon command makes the C compiler execute on a machine called floppsie all input/output has to be rerouted tofloppsie in order that the compilation will work. telnet -the ability to login into another machine NFS -Network file system. RPC -Remote procedure call package (1984)

26 normally uses thexdr facility and it is an interprocess communication mechanism that allows programmers to write client-server based applications

27 UNIX networking facilities slide 27 sockets -introduced in 1982 with 4.1 BSD. Another interprocess communication mechanism (low level) which allows one process to communicate with another process. These processes maybe on the same machine or on opposite sides of the world. Sockets consist on five primitives:socket,bind,listen, connect andaccept today hundreds of client/servers use above primitives (pop2, pop3, imap, smtp), news, R commands, lpd, nfs, ftp, telnet, news, finger, cvs, http

28 Networking examples slide 28 pop2 pop3 imap smtp news Rcommands lpd nfs ftp telnet finger cvs http client client server client discussion groups remote commands networked printing networked file system file transfer protocol remote login user description & profile networked source code access hypertext protocol

29 Networking examples slide 29 examine the file/etc/services for a full description

30 Atypical UNIX network slide 30 UNIX is an open system many vendors: SUNOS, ULTRIX, AT&T UNIX SVr4, GNU/Linux, BSD386 some free! (Source code, compilers, debuggers, editors, network applications etc) its popularity in the early days was due in part to its no cost availability this has now turned full circle with the advent of GNU/Linux and BSD386 open system as interfaces to applications, client/servers are published important to note amajor difference between UNIX and other operating systems (Novell) is that clients and servers onunix look almost identical

31 possible that they are the same except for some files in/etc (configuration files) Atypical UNIX network slide 31 client/server distinction occurs at the user level not the system level

32 Atypical UNIX network slide 32 workstation local disk workstation local disk ethernet and IP workstation local disk configuration is very flexible

33 Unix configuration slide 33 could be a disadvantage as system administrators have to fight their way through many different options for various components just to perform simple operations it is possible that all machines are configured identically (except for their hostname, IP address etc) could configure the network operating systems so that all local disks can be seen by all machines

34 Atypical UNIX network slide 34 alternatively machines could be very different (ie could build distinct client/server model similar to Novell) one machine has the common file system one machine has the system executables

35 Disadvantages slide 35 complexity flexibility requires system administrators be disciplined otherwise overall system may become unmanageable when more machines are added harder to configure than other operating systems getting better (Solaris, GNU/Linux brands often use Gtk) but complexity still remains few graphical tools to identify installation errors however the problems are being addressed

36 NFS configuration example slide 36 can network file systems between different architectures little endian, big endian - it makes no difference NFS design is built from three main parts: a protocol, a server and a client nfs client nfs protocol nfs server

37 NFS configuration example slide 37 components fit into the ISO-7 layer model Level 7 Application NFS Level 6 Presentation XDR Level 5 Session RPC Level 4 Transport TCP UDP Level 3 Network IP Level 2 Data Link Logical Level 1 Physical Link Ethernet

38 NFS example slide 38 suppose we have two machines floppsie mcgreg / / bin home usr home usr bin X386 fred bin include our task is to allow machinefloppsie see directory /home/gmulley

39 NFS example slide 39 tellfloppsie where this directory resides edit/etc/fstab on floppsie and add the following line mcgreg:/home /home/mcgreg nfs (remote fs mount point type mount options)

40 NFS example slide 40 need to tell machinemcgreg that/home is exported and that floppsie is allowed to see it edit/etc/exportfs onmcgreg and add the following line /home / (rw) floppsie(rw)

41 Plan 9 from Bell Labs slide 41 MOVIE GOOFS AND INTERESTING FACTS LIST Version 3.9 (October 1993) Acollection of film flubs, cinematic slip-ups, movie mistakes, Hollywood howlers, etc etc. Compiled by Murray Chapman and Michael Gaines (render1@teak.njit.edu) Found by the users of USENET. #Plan 9 from Outer Space -Bela Lugosi died two days into the shooting of this turkey. He was replaced by the director s wife s chiropractor, who was significantly taller than Lugosi, and played the part with a cape covering his face. -Internationally recognised as the worst movie ever made.

42 Plan 9 from Bell Labs slide 42 Plan 9 from Bell Labs is a network operating system and it is composed of separate machines acting as CPU servers file servers terminals it is built from distinct specialised components rather than similar general purpose components high performance general purpose components cost a lot of money workstations are too slow and IO bound for fast compilation too expensive just to be used as a windowing system

43 Plan 9 from Bell Labs slide 43 Plan 9 approach is based on distributed specialisation would better address issues: cost effectiveness maintenance performance reliability and security

44 Plan 9 from Bell Labs slide 44 radical steps taken build a completely new system Ccompiler, operating system, network software, command interpreter, window system and terminal! authors credentials? Rob Pike -famous for many ofunix concepts during the early years Ken Thompson - wrote the first C compiler. Then worked on translating UNIX from assembler to C AT&T - birth place of UNIX - so wealthy that the USA divided AT&T into two because of its power

45 Plan 9 slide 45 different from other research and commercial operating systems. It lacks features often associated with other research network operating systems process migration lightweight processes distributed file caching personalised workstations support for X windows

46 Plan 9 slide 46 built around CPU servers (large multiprocessors) file servers terminals (bitmap screen, mouse & computer running a window system 8 1/2) sharing of computing and storage services provide a sense of community for a group of programmers cost effective simplifies management and administration

47 CPU Servers slide 47 in 1990 the production CPU server was a Silicon Graphics Power Series machine with four 25 MHz MIPS processors 128 Mbyte memory no disk 20 Mbyte/sec DMA interface to the file server Ethernet network interface

48 CPU Servers slide 48 operating system has conventional process model using fork exec primitives new processes are run on a new processor whenever possible (similar to Amoeba in this respect) processors similar to disk blocks buy more disk space when you run out buy more processors as well CPU servers have no local disk

49 File Servers slide 49 Plan 9 file server is a modest two processor 64 Megabyte of memory 600 Megabytes of hard disk 300 Gigabyte juke box of write-once optical disk (WORM) it interfaces to the CPU servers through the 20 Mbyte/sec DMA links

50 File Servers slide 50 file server interfaces to other terminals through conventional networks file server presents a file system to its clients not an array of disk drives different from Windows in this respect. cfa: b: c: d: e: f:

51 File Server (continued) slide 51 file system actually resides on WORM the modest 600 Mbyte drive only acts as a cache file server client Mbyte Mbyte Gbyte RAM disk worm

52 File Server Security and Backup slide 52 once a day the file tree is frozen and split old files can be accessed through normal file operations /1990/0401/usr/rob/doc/plan9.ms this scheme permits users to compare files against old versions using traditional tools applied to a normal file system permissions are the same security remains intact

53 Plan 9 Network Structure slide 53 Ethernet 20 Mb/s DS-1 DMA GNOT File Server CPU Server

54 GNOT terminal slide 54 specialist terminal whose specification is: 25 MHz x colour display keyboard and mouse 4..8 Mbytes of memory 2Mbit/sec packet switched network connects a GNOT tothe CPU servers terminal does not handle compilation might run a text editor though

55 Plan 9 GNOT slide 55 cheap enough so that each person can have 2 1atwork 1athome

56 Plan 9 code size statistics slide 56 window software written from scratch 81/2 source is 60K bytes including source for xterm equivalent roughly functionality of X windows (no colour) 3860 lines of code and compiles in 10 seconds compare to X windows! Plan 9 is small! the kernel is 3647 lines of C code it takes 10 seconds to compile and 10 seconds to load

Network File System (NFS)

Network File System (NFS) slide 1 References: Douglas Comer, Internetworking with TCP/IP, p430-432 Simson Garfinkel and Gene Spafford, Practical UNIX and Internet Security, p605-631 NFS is used by GNU/Linux