05/01/2017 Today Comp 104: Operating Systems Concepts • OS evolution Processes • Introduction to processes • OS structure Management Scheduling & Resource Allocation 1 2 Evolution of OS Multiprogramming • Largely driven by desire to do something useful when a • Load several programs into memory program cannot continue (maximise throughput ) simultaneously, all sharing single CPU • Early systems: –‘Job’ loaded from punched cards or tape, output to printer – Job may include loading compiler, assembler, linker, data etc. • When running program cannot continue – CPU idle for much of the time (e.g. waiting for I/O), switch to another • Batch systems: – Job passed to human operator – Operator groups jobs into batches with similar characteristics, e.g. all programs using same compiler • Hence, I/O and computation overlap – More efficient use of resources 3 4 1 05/01/2017 Multi-Access (Time-Sharing) Question • The following two statements describe the performance of two programs • An extension of multiprogramming (where the computation and input/output could be interleaved): – A performs a total of 20 seconds of computation and 15 seconds of input/output. – B performs a total of 30 seconds of computation and 10 seconds of I/O • CPU is switched rapidly between • Which of the following are true? processes to give illusion of uninterrupted I. It will take up to 50 seconds to run A and B sequentially II. It will take up to 75 seconds to run A and B sequentially execution in parallel (multitasking) III. Using multiprogramming, the shortest time to execute both is 50 seconds – users can interact with programs IV. Using multiprogramming, the shortest time to execute both is 40 seconds – users see their own ‘virtual machine’ a) I and III b) l and lV Answer: c – resources (printers, disks etc.) are shared, but c) ll and lll •If run sequentially, A needs to finish before B d) ll and lV can begin, therefore II is true. this is largely transparent •With multiprogramming, I/O for one process can e) None of the above take pave whilst the computation takes place for another. Therefore III is true 5 6 Operating System – An Abstract Implications View • Need to decide which programs to load from User Command Interface disk into memory (job scheduling) • Need to decide which program to execute next (CPU scheduling) Processor Manager • Consider disk space as extension of main memory (virtual memory) Memory Manager • Memory allocation Device Manager • Disk/file allocation • Protection/security File Manager 7 8 2 05/01/2017 Processes Process Characteristics • A program is a representation of an • Process characteristics: algorithm in some programming language; – Requires space in memory where it resides during i.e. it is static execution – During its execution it may require other resources • A process refers to the activity performed such as data files or I/O by a computer when executing a program; i.e. it is dynamic – It passes through several states from its initial creation to its completion within the computer system (more details on these states to come in later • A process is created when a program or lectures) command is executed 9 10 Processes Processes • A process needs resources, such as CPU time, • The Processor Manager is responsible for memory, files and I/O devices, to accomplish its task. overseeing the following activities in relation to process management: • These resources are allocated either when the – Creation and deletion of both system and user program is created, or when it is executing. processes • Operating-system processes execute system – Scheduling processes code and user-processes execute user code – Provision of mechanisms for synchronisation – All these processes could potentially execute and communication of processes concurrently – Deadlock handling for processes 11 12 3 05/01/2017 O.S. Structure Command Interpreter • Often consists of: • Accepts and runs commands specified by user – A central nucleus or kernel – Hence provides user’s view of OS • resides permanently in memory • May be graphical, e.g. Windows • performs low-level, frequently needed activity • May be textual, e.g. UNIX shell – A set of processes – bash, ksh, csh • may be system level or user level – Some commands built into shell, others loaded from – processes interact with kernel via system separate executable files calls – Shell also has sophisticated control structures such • e.g. create process, run program, open file as loops, if-statements and procedures – kernel and system level processes may operate in privileged mode 13 14 Process States State Changes • Running – on a uniprocessor machine, only one process can terminated be executing at any time new admitted – may be interrupted at end of time-slice if no I/O interrupt exit requests or system calls performed ready running • Ready dispatch – refers to a process that is able to run, but does not I/O or event I/O or event wait currently have the CPU completion • Waiting(Blocked) blocked – refers to a process that is unable to continue, even if granted the CPU 15 16 4 05/01/2017 Question Process Descriptors • A running process makes a system call to read data from a file. Which process state should it enter next? • For each process, the OS kernel maintainers a descriptor or Process Control Block (PCB) a) New • PCB contains info like b) Ready c) Running – unique process ID d) Blocked – user ID of process owner e) Terminated – process state – position in memory – accounting stats. (time used etc.) Answer: d Blocked; it may take some time before the file system can read the – resources allocated (open files, devices, etc.) file (e.g. on a networked file store), so the process is blocked until – register values (process counter, etc) the data is available. 17 18 Context Switch PCBs and Queuing • When a process is interrupted • The PCB of each process is updated as the – all current state information (including program process progresses from the start to the end of counter and other registers) is saved into PCB its execution – PCB is put into a queue • may have several, e.g. for different devices – the kernel may do some of its own work • Queues use PCBs to track the processes’ • e.g. handling a system call progress through the system. The PCBs are – the PCB of a process from the ready queue is linked to form queues: selected, and its context restored – ‘Ready queue’ linking the PCBs for every ‘ready’ • Whole context switch is an expensive process overhead – ‘New queue’ linking the PCBs for processes just entering the system – hardware support may help • e.g. multiple register sets 19 20 5 05/01/2017 PCBs and Queuing Queuing terminated new • Processes that are ‘blocked’ are linked together admitted by ‘reason for waiting’ interrupt exit ready running – PCBs for these processes are linked into several I/O or event dispatch queues completion • e.g. those waiting for I/O on a specific disk drive are linked together, those waiting for a printer are linked in a different blocked queue I/O request Disk I/O queue • All queues need to be effectively managed in an Printer I/O queue order that is determined by the process scheduling policies and algorithms Other I/O queue 21 22 Inter-Process Communication Signals • A process can usually be terminated by typing • Inter-Process Communication (IPC) CTRL-C mechanisms allow processes to talk to each – Actually sends a signal to process other – Process responds by aborting • IPC useful when processes working together • Signals can be sent from one process to another (cooperating processes) – signal() system call • Signals can be sent from the command line – synchronisation and/or passing data using kill command • For example in UNIX: – Format: kill -<signal> <pid> – signals – e.g. kill -9 12345 sends signal 9 (kill signal) to process 12345 – pipes – sockets 23 24 6 05/01/2017 Responding to Signals Pipes • A receiving process can respond • The UNIX command ‘wc –l file’ counts the number to a signal in three ways: Example kill signals of lines in file – Perform default action (e.g. abort) 1 HUP (hang up) • If we just type ‘wc –l’ we don’t get an error – Ignore the signal 2 INT (interrupt) 3 QUIT (quit) • Instead, data is read from standard – ‘Catch’ the signal; i.e. execute a 6 ABRT (abort) Common wc flags 9 KILL (non-catchable, non- input (keyboard by default) designated procedure ignorable kill) -l number of lines 14 ALRM (alarm clock) – Similarly for output files and standard -w number of words 15 TERM (software output (screen) -c number of characters • The ‘kill’ signal (signal 9) cannot termination signal) By default, all three be caught or ignored • The pipe symbol ‘|’ attaches the stats are displayed. – Guaranteed way to stop process standard output of one program to Flags state what stats appearL the standard input of another, e.g. 25 who | wc -l 26 Client-Server Examples • The following are examples of common servers: – Web server: accessed by client’s web browser – Mail server: retrieving and sending emails to clients – File server: holding documents to be accessed by clients – Database server: providing database services to clients, e.g. customer database, stock databaseL – etc 27 7.