I/O topics


  device classification

    behavior - input, output, or storage
    partner - person or machine
    data rate


  reliability/availability

    MTTF - mean time to failure
    annual failure rate (AFR) - percent of devices expected to fail in a year

    MTTR - mean time to repair
    availability = MTTF/(MTTF+MTTR)

    failure causes - HW, SW, operator error

    fault avoidance - better design/construction
    fault tolerance - use of redundancy, e.g., RAID
    fault forecasting - preventative maintenance/replacement

    faclt detection
    fault diagnosis
    repair


  disk - see CPSC 231 notes

    three components of access time
      1) seek
      2) rotational latency
      3) transfer time (R/W)

    controller overhead

    caches in controllers

    AFR = (number_of_drives * usage_time_per_year)/MTTF


  bus

    low cost, shared communication medium
    limit on bandwidth

    synchronous - must be short to be fast, all devices run at same speed
    asynchronous - more flexible, can be longer, protocol overhead

    trend of moving to high-speed serial point-to-point connections
      (e.g., USB)


  interrupts - see CPSC 231 notes


  DMA - see CPSC 231 notes