Chapter One -- Introduction -- Lecture Notes

An OS performs as an intermediary between the user and the hardware, functioning as a virtual computer that makes the use of the hardware more convenient and/or efficient.


• 1.0 Chapter Goals
  
  
  • Describe basic organization of computer systems
  • Provide tour of major components of OS's
  • Overview the many types of computing environments
  • Explore open-source Operating Systems
  
  
• 1.1 What Operating Systems Do
  
  
  • Provide An Interface above the Level of the Bare Machine
  • Make the Computer More Efficient
  • Make Computer Use More Convenient
  • Controls and Coordinates Use of the Hardware - esp. I/O - prevents error and misuse.
  • The OS is a manager and allocator of resources: CPU time, storage, devices ... There is competition among processes for resources.
  • Provide Functions Commonly Needed by Application Programs
  • Ease of use is a major goal for a PC operating system.
  • Optimizing resource utilization is more important for operating systems on Workstations and mainframes.
  • OS features that help conserve battery life are important on hand-held computing devices.
  • IMPORTANT: In this course, we will think of the OS as the Kernel - the part of the system that is always on the list of ACTIVE processes - READY to run anytime it can be allocated time on a CPU.
  
  
• 1.2 Computer-System Organization
  
  
  • The typical system consists of one or more general-purpose CPU's and special purpose controllers sharing a common bus, including a memory controller. These units are able to operate in parallel, though they compete for the memory. The memory controller arbitrates.
  • A firmware-resident routine runs at boot time to perform initializations and load & execute the OS.
  • Hardware devices can interrupt the CPU - a signal sent along the system bus.
  • A process can interrupt the CPU - by executing a special instruction - also by making an error.
  • The computer is designed in such a way that an interrupt automatically causes suspension of the currently running process and transfer of execution to some predetermined section of code - typically a part of the operating system - an interrupt handler.
  • When an interrupt occurs, everything that will be needed to restart the currently running process must be saved. The interrupt automatically does some of the saving - for example the saving of the address of the current instruction. The interrupt handler code may perform other "context-saving" actions.
  • CPU's go through an automatic instruction fetch, decode and execute cycle.
  • The CPU can access only programs and data that are in main memory. Permanent copies reside in files on secondary storage. The smallest unit of addressable primary memory is usually a byte.
  • Memory hierarchies consist of many levels with differing speeds, costs, sizes and volatility.
  • Typically methods for controlling peripheral devices are quite diverse and complex. Operating systems have "device driver" code - drivers corresponding to each type of device controller. The job of the device driver is to receive simple high level commands from the OS, translate the commands, and communicate with the device to achieve the desired effects.
  • Typically drivers have to load bit patterns into registers within the devices in order to "command" the device to perform actions.
  • Typically, when a device finishes performing a data transfer it notifies the OS by sending an interrupt.
  • With Direct Memory Access (DMA) a controller can transfer an entire block of data between device and primary memory, and interrupt the CPU only once, after the entire block has been transferred.
  
  
• 1.3 Computer-System Architecture
  
  
  • OS's must rise to the challenge of operating various kinds of multiprocessor systems efficiently: SMP's, multiple cores, blade servers & clustered systems.
  • A Multiprocessor System (tightly-coupled system) has two or more general-purpose CPU's that share a bus, and often the clock, main memory and peripherals. Typically a multiprocessor with N CPU's is cheaper than N single-CPU computers and does almost as much work in as little time. If designed properly, a multiprocessor will exhibit "graceful degradation" and "fault tolerance."
  • A multiprocessor may be asymmetric, but most are symmetric nowadays.
  • There may be multiple CPUs having their own memory controllers - a CPU may be able to access some parts of primary memory more quickly than others - non-uniform memory access (NUMA).
  • Multiple cores are essentially multiprocessors on a chip. These tend to have the advantages of quicker communication and lower power requirements than multiprocessors constructed from multiple single-core chips.
  • Clustered Systems are like multiprocessors. They exhibit graceful degradation but are less tightly-coupled, consisting of computers closely connected on some type of high-speed medium, often sharing secondary storage, but not main memory or clock.
  • There are asymmetric and symmetric clustered systems.
  • The forms of parallel processing and sharing found on multiprocessors and clustered systems require sophisticated forms of synchronization and data "locking" to prevent incorrect results from "conflicting operations".
  • Clustered systems often make use of storage-area networks (SANs). SANs provide a way for multiple computers to share multiple secondary memory devices (typically disk drives).
  
  
• 1.4 Operating-System Structure
  
  
  • Multiprogramming was a development pushed into existence by the need for better utilization - the need to keep the CPU and I/O channels of an expensive computer busy, so the computer will do more work per unit time. A "QUANTUM LEAP" in sophistication of the OS was required -- e.g. context switching, memory management, scheduling, concurrency management, protection, and security.
  • Time-sharing was a "natural" follow-on to multiprogramming that allowed a return to "interactive" computing. Processes associated with different users are switched very quickly so that every user gets rapid response to every command.
  • Time-sharing tends to make greater demands for swapping and virtual memory.
  • Traditional time-sharing requires implementing a file system -- not the same as just the ability to write and read disk sectors.
  
  
• 1.5 Operating-System Operations
  
  
  • A modern operating system is interrupt driven. It is interrupts and traps alone that allow the operating system to regain access to the CPU.
  • Hardware Protection mechanisms: The hardware and OS should prevent processes from harming the system and each other. The OS must be protected. We must be concerned about illegal I/O, illegal memory access, and "hogging" of the CPU. Tools: traps -- dual mode operation -- memory protection -- base and limit registers -- privileged instructions -- timers.
  • In your mental model of how an OS works, do you see the OS as a part of the system that is always executing and always monitoring user processes? Typically "the OS" is NOT REALLY running all the time. In fact this would make it impossible for user processes ever to execute on a uniprocessor. An OS can employ hardware protection mechanisms to ensure that it will not lose control of the system, even during periods when the OS is suspended.
    
    
  • User processes make requests for services from the OS by making system calls, typically implemented by executing an instruction that triggers a trap.
  • Test your understanding: Which is responsible for changing the mode bit from 0 (kernel mode) to 1 (user mode): the hardware or the operating system? Which is responsible for changing the mode bit from 1 to 0: the hardware or the operating system?
  
  
• 1.6 Process Management
  
  
  • A process is a program in execution. An OS has to schedule, create, delete, suspend, resume, and synchronize processes.
  • Processes compete for resources. They have to request them from the OS and wait until the OS makes them available.
  • An OS must provide mechanisms by which processes can communicate.
  
  
• 1.7 Memory Management
  
  
  • Main memory consists of a sequence of 'words' - each of which has an address. We use the term 'word' to refer to whatever is the smallest addressable unit of memory. On many computers, the word is a byte, but it could be something else - e.g. a 32- or 64-bit word.
  • A computer cannot read directly from secondary memory like it can from main memory. Because of the way computers are designed, in order for a CPU to execute an instruction, the instruction has to be in main memory.
  • The OS has to allocate memory to processes and deallocate it. It has to keep track of which portions of memory belong to which processes
  
  
• 1.8 Storage Management
  
  
  • Operating Systems are responsible for implementing file systems on secondary storage: e.g. creating and deleting files & directories, and allocating/deallocating storage space on disk.
  • It is the operating system that creates, out of collections of sectors scattered over secondary memory, what we perceive as a highly organized system of directories and files.
  • The OS has to create and delete files and directories, support primitive operations for manipulating files (e.g. open, seek, read, write and close), map files onto locations in secondary storage, and provide a means for backing up files
  • The OS has to manage free space on disks, allocate storage to files, and may be responsible for scheduling disk operations.
  • Various kinds of caches are important for performance. Hardware-only caches are outside the control of the OS.
  • The copying of data between primary and secondary memory is usually under the control of the OS.
  • The OS has to solve consistency and coherence problems, because copies of the same data item may exist concurrently at different levels of the memory hierarchy, and in the caches of more than one CPU. When one process changes one copy, how will the other copies be updated so that every process looking at any copy of the data will see the same thing?
  • One of the jobs of the OS is to make all the various I/O devices appear to have the same familiar types of controls, when in fact there are many peculiarities in how different devices are designed to be controlled. Operating Systems provide high-level commands to operate devices, as well as device drivers that interpret the high-level commands and deal with the device-specific commands that are required. So, for example, an application can ask the OS to get the next character from an input device, and the OS will tell the device driver to get the next character, and then the device driver will issue all the peculiar commands to the device that are needed to get the next character.
  
  
• 1.9 Protection and Security
  
  
  • Although the difference between computer protection and security has become blurred, our text makes an effort to preserve a clear distinction between the two.
  • An OS must perform "gatekeeper" functions - making sure that users provide proper credentials before being given access to system resources. This is "protection". The doorman at an expensive hotel and the sentry at a military facility are doing protection work.
  • gatekeepers and sentries certainly cannot adequately prevent all failures, damage or attacks that could hinder the correct operation of the facility. Besides protection, the designer of an OS must be concerned with "security" - insuring that resources are properly used. It is security work when casino owners watch customers gamble. The bouncer in a night club and the teacher on yard duty do security work.
  
  
• 1.10 Kernel Data Structures
  
  
  • Among the data structures employed by operating systems are:
    • Arrays
    • Lists
    • Stacks
    • Queues
    • Trees (general, binary, unbalanced, balanced)
    • Hash functions and tables
    • Bitmaps
  
  
• 1.11 Computing Environments
  
  
  • Traditional Computing - at the office, roughly chronologically: terminals attached to mainframes, networked PCs, file- and print-servers, laptops, portals providing web access to internal servers, network terminals, handheld devices sync'd to PCs. At home slow modem connections evolved into small home networks with broadband network connections.
  • Organization portals now provide web access to internal servers. Mobile devices are integrated via wireless computer networks and cellular data networks.
  • Mobile devices implement location-based services and have applications that rely on input from accelerometers. Mobile devices are very portable and feature-rich, but have small displays. They also have less computing speed compared to larger computers, and less primary and secondary memory capacity.
  • A distributed system is basically a set of computers in communication via a network.
  • Network technologies vary. Local-area, wide-area, metropolitan-area, and personal-area are some of the classifications.
  • A network operating system provides network services to the hosts in a network. Hosts function independently in a network operating system.
  • A distributed operating system takes the idea of a network operating system to a higher level. It is operating system software that makes a distributed system function as if it were a single computer.
  • Client-Server Computing - client processes make requests for services offered by server processes, typically file service and compute service.
  • Peer-to-Peer Computing - exploits potential of having both client and server processes on a node, and schemes to avoid server bottlenecks through better distribution of clients across the set of available servers. There may be some use of broadcasting or centralized service.
  • Virtualization allows guest operating systems to run within a host operating system - useful if one wishes to test one or more operating systems on a single computer, or if one wishes to run programs that are designed for a certain non-native operating system, for example, running a Windows program on a Mac.
  • Cloud Computing is delivery of computing power, storage, and/or programs as a network service. There are public and private clouds, and hybrids. Word processors, spreadsheets, database service, and backup storage are among the resources one can access.
  • Real-Time Embedded Systems typically have primitive OSs and primarily perform monitoring and management of devices. A real-time system has to deal with time constraints imposed by the external environment. (By the way, this involves more than just being quick. As an example think about a robot whose job is to open a heavy door for you, and then close it behind you. If you take a run at the door, you don't want the robot to be too late opening it or too early closing it!)
  
  
• 1.12 Open Source Operating Systems
  
  
  • Open source software is made available in source form, not just binary code. This enables programmers to modify it to increase its usefulness to them. They may find and fix bugs made by programmers who worked on the code before them, and pass along the improved software to other users. Users of open-source software tend to form associations and give each other help and support, as well as make suggestions for improvement of the software.
  • History - open-source software was in vogue in the early days of computing, became less prevalent with the corporatization of computing, but has come back into quite a lot of prominence, notably under the influence of the Free Software Foundation, founded by Richard Stallman.
  • Linux - the original kernel was designed by Linus Torvalds (probably with considerable influence from Andrew Tannenbaum's Minix). It was combined with compilers and utilities produced by Stallman's GNU project. Many distributions now exist. Thousands of programmers have worked on Linux (open) source code.
  • BSD Unix - The flavor of unix developed at UC Berkeley - was tied to AT&T unix for some time & required a license, but now is free of AT&T code. There are many descendants, including the Darwin kernel component code used in the Mac OS X.
  • Solaris - originally the OS of Sun Microsystems was based on Berkeley unix. Sun migrated to the AT&T based Solaris and eventually open-sourced most of the code it developed. Some of Solaris is still closed source.
  • Utility - one might say that open-sourcing is a movement that is helping to increase participation in software development.