Storage Structure and Computer Booting Process

Storage Structure

• Storage systems are organized in a hierarchy based on speed, cost, volatility, and size/capacity
• Registers are the fastest and most expensive memory, while magnetic tapes are the slowest and least expensive
• Volatile memory loses its data when power is lost (Registers, Cache, Main Memory)
• Non-volatile memory retains data even without power (Solid State Drives, Magnetic Disks, Optical disks, Magnetic Tapes)

Computer Startup/Booting Process

• Booting is a process that loads the OS when the user turns on the power
• The bootstrap program is loaded at power-up or reboot
• The bootstrap program is stored in ROM or EPROM (firmware)
• Bootstrap program initializes all system aspects
• Bootstrap program loads the OS kernel and starts execution
• Boot process steps:
• User presses the start button, creating an electric signal
• Power supply sends wake up signals to the CPU
• CMOS Battery powers the CMOS Memory
• Power on Self Test (POST) runs
• If the POST fails the error is displayed
• Master Boot Record(MBR) is loaded
• OS starts

Computer-System Architecture

• Single-processor systems contain one CPU for general-purpose instructions
• Some single-processor systems also have special-purpose processors for tasks like disk, keyboard, DMA, and graphic control
• OS may manage special processors, like disk controllers, to schedule algorithms
• Multiprocessor systems have two or more processors working together
• Multiprocessor systems share computer bus, system clock, memory, and I/O devices
• Multiprocessor systems are also called parallel or multicore systems

Multiprocessor Systems: Advantages

• Multiprocessor systems increase thoughput to complete more work in less time
• Increased speed up in n processor system is not n due to overhead
• Multiprocessor systems are more economical due to shared I/O devices, memory, and power
• Multiprocessor systems increase reliability because failure will not stop the entire system

Multiprocessor Systems: Reliability

• Graceful degradation allows the system to maintain operations proportional to surviving parts
• Fault-tolerant systems continue functioning during component failures
• Fault-tolerant systems require failure detection, diagnoses, and correction
• Fault-tolerant systems may duplicate hardware/software to execute tasks in parallel

Multiprocessor Systems: Types

• Asymmetric multiprocessing uses a master processor to schedule and allocate work to slave processors
• Each slave processor waits for instructions from the master or has a predefined task
• Often used in extremely large systems
• Symmetric multiprocessing (SMP) does not use a master-slave model
• Each processor runs an identical copy of the OS, has its own registers and cache
• SMP shares the same memory
• SMP ensures that many processes can run without significant performance deterioration
• SMP often needs load balancing to improve performance
• Symmetric and asymmetric configurations can result from hardware or software design

Clustered Systems

• Clustered systems consists of multiple CPUs, like multiprocessors
• Clustered systems are individual systems or nodes
• Each node can be a single processor or multicore
• Clustered system share storage and communicate via LAN
• Clustered systems offer high-availability service

Interrupts

• Interrupts are signals sent to the CPU by external I/O devices
• Interrupts tell the CPU to stop its current activities and execute the OS
• There are three types of interrupts:
• Hardware interrupts are generated by hardware devices needing attention
• An example of hardware interupts are keystrokes on the keyboard completing a data transfer
• Software interrupts are generated by programs requesting a system call
• Traps are generated by the CPU itself to indicate an error requiring OS assistance

Interrupt Handling

• An interrupt suspends one program's execution to begin another
• Separate OS code segments handle each interrupt type
• The code that handles interrupts is called an Interrupt Service Routine
• Each I/O device has an associated memory location called an Interrupt Vector
• The Interrupt Vector contains the ISR's address

Interrupt Handling Process

• When an interrupt (or trap) occurs, hardware transfers control to the OS
• The OS preserves the CPU's state by storing registers and the Program Counter
• ISRs determine what action should be taken for each type of interrupt
• Intel Pentium interrupt vector table lists the vector number and description

Operating System Structure

• General-purpose OS is a large program that may be structured in various ways
  • Simple structure (MS-DOS)
  • More complex structure (UNIX)
  • Layered structure (an abstraction)
  • Microkernel (Mach)

Simple Structure: MS-DOS

• MS-DOS was written to provide the most functionality in the least space
• MS-DOS is not divided into modules
• MS-DOS has some structure, but its interfaces and functionality levels are not well separated

Non-Simple Structure: UNIX

• The original UNIX operating system had limited structuring due to hardware limitations
• UNIX OS consists of two separable parts:
• Systems programs
• The kernel consists of everything below the system-call interface and above physical hardware
• The kernel provides file system, CPU scheduling, memory management, and other operating-system functions

Layered Approach

• Layered approach divides the OS into levels, stacked on top of one another in layers
• The bottom layer (layer 0) is the hardware
• The highest layer (layer N) is the user interface
• Layers use operations/services of only lower-level layers by using MODULARITY

Microkernel System Structure

• Microkernel moves most of the kernel into user space
• Mach is an example of microkernel
• Mac OS X kernel (Darwin) is partly based on Mach
• Communication takes place between user modules using message passing
• Benefits of microkernel:
• Easier to extend the OS and port to new architectures
• More reliable and secure
• Detriments of microkernel:
• Performance overhead of user space to kernel space communication

Modules

• Modern operating systems implement loadable kernel modules
• This approach uses object-oriented programming
• Each core component is separate and talks to others over known interfaces
• Every module is loadable as needed within the kernel
• Modules are similar to layers but are more flexible (Linux, Solaris)