CS 1550: Introduction to Operating SystemsClass outlineOverview: Chapter 1What is an operating system?Operating system timelineFirst generation: direct inputSecond generation: batch systemsStructure of a typical 2nd generation jobSpoolingThird generation: multiprogrammingTimesharingTypes of modern operating systemsComponents of a simple PCCPU internalsStorage pyramidDisk drive structureMemoryAnatomy of a device requestOperating systems conceptsProcessesInside a (Unix) processDeadlockHierarchical file systemsInterprocess communicationSystem callsMaking a system callSystem calls for files & directoriesMore system callsA simple shellMonolithic OS structureVirtual machinesMicrokernels (client-server)Metric unitsChapter 1CS 1550:Introduction to Operating SystemsProf. Ahmed [email protected]://www.cs.pitt.edu/~amer/cs1550Chapter 12CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)Class outlineIntroduction, concepts, review & historical perspectiveProcessesSynchronizationSchedulingDeadlockMemory management, address translation, and virtual memoryOperating system management of I/OFile systemsSecurity & protectionDistributed systems (as time permits)Chapter 13CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)Overview: Chapter 1What is an operating system, anyway?Operating systems historyThe zoo of modern operating systemsReview of computer hardwareOperating system conceptsOperating system structureUser interface to the operating systemAnatomy of a system callChapter 14CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)What is an operating system?A program that runs on the “raw” hardware and supportsResource AbstractionResource SharingAbstracts and standardizes the interface to the user across different types of hardwareVirtual machine hides the messy details which must be performedManages the hardware resourcesEach program gets time with the resourceEach program gets space on the resourceMay have potentially conflicting goals:Use hardware efficientlyGive maximum performance to each userChapter 15CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)Operating system timelineFirst generation: 1945 – 1955Vacuum tubesPlug boardsSecond generation: 1955 – 1965TransistorsBatch systemsThird generation: 1965 – 1980Integrated circuitsMultiprogrammingFourth generation: 1980 – presentLarge scale integrationPersonal computersNext generation: ???Systems connected by high-speed networks?Wide area resource management?Chapter 16CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)First generation: direct inputRun one job at a timeEnter it into the computer (might require rewiring!)Run itRecord the resultsProblem: lots of wasted computer time!Computer was idle during first and last stepsComputers were very expensive!Goal: make better use of an expensive commodity: computer timeChapter 17CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)Second generation: batch systemsBring cards to 1401Read cards onto input tapePut input tape on 7094Perform the computation, writing results to output tapePut output tape on 1401, which prints outputChapter 18CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt) $END $RUN $LOADStructure of a typical 2nd generation job $FORTRAN $JOB, 10,6610802, ETHAN MILLERFORTRANprogramData forprogramChapter 19CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)SpoolingOriginal batch systems used tape drivesLater batch systems used disks for bufferingOperator read cards onto disk attached to the computerComputer read jobs from diskComputer wrote job results to diskOperator directed that job results be printed from diskDisks enabled simultaneous peripheral operation on-line (spooling)Computer overlapped I/O of one job with execution of anotherBetter utilization of the expensive CPUStill only one job active at any given timeChapter 110CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)OperatingsystemThird generation: multiprogrammingMultiple jobs in memoryProtected from one anotherOperating system protected from each job as wellResources (time, hardware) split between jobsStill not interactiveUser submits jobComputer runs itUser gets results minutes (hours, days) laterJob 1Job 2Job 3MemorypartitionsChapter 111CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)TimesharingMultiprogramming allowed several jobs to be active at one timeInitially used for batch systemsCheaper hardware terminals -> interactive useComputer use got much cheaper and easierNo more “priesthood”Quick turnaround meant quick fixes for problemsChapter 112CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)Types of modern operating systemsMainframe operating systems: MVSServer operating systems: FreeBSD, SolarisMultiprocessor operating systems: Cellular IRIXPersonal computer operating systems: Windows, UnixReal-time operating systems: VxWorksEmbedded operating systemsSmart card operating systemsSome operating systems can fit into more than one categoryChapter 113CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)Components of a simple PCHard drivecontrollerVideocontrollerMemoryUSBcontrollerNetworkcontrollerOutsideworldCPUComputer internals(inside the “box”)Chapter 114CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)ExecuteunitExecuteunitExecuteunitExecuteunitBufferFetchunitDecodeunitFetchunitDecodeunitFetchunitDecodeunitCPU internalsPipelined CPU Superscalar CPUChapter 115CS 1550, cs.pitt.edu (originaly modified by Ethan L. Miller and Scott A. Brandt)Access latency1 ns2–5 ns50 ns5 ms50 sec< 1 KB1 MB256 MB40 GB> 1 TBCapacityStorage pyramidRegistersCache (SRAM)Main memory (DRAM)Magnetic diskMagnetic tapeGoal: really large memory with very low latencyLatencies are smaller at the top of the hierarchyCapacities are larger at the bottom of the hierarchySolution: move data between levels to create illusion of large memory with low latencyBetterBetterChapter 116CS