---

Techniques to speed up access to data stored on hard disk

---

• General approaches to speed up access to data stored on disks
  • Placement:

    Place sectors of blocks (= related) in same cylinder Allocate blocks of the same file in same or nearby cylinders (Disk de-fragmentation)

    ---

    ---

  • Hardware:

    Use multiple smaller disks instead of one large disk Advantages: Multiple independent disk heads can have one disk seeking while another transfering data (Overlap operation Multiple disks can higher total transfer rate

  ---

  ---

  ---

  ---

• Striping: a storage technique that uses multiple disks
  • Striping:

    Striping = storing the blocks of the same file over multiple disks

    Graphically:

    ---

  • Reported experience:

    With 4 disks, striping achieves an average speed up of 3 times the speed of 1 (big) disk (Assuming that the disk controller has a high enough transfer speed of 4 disks)

  ---

  ---

  ---

  ---

• Achieving better performance through scheduling
  • Fact:

    There are many disk requests at any one moment in time:

    ---

    ---

  • Service policy:

    Service policy: in which order do you process (= service) the pending requests ?

    ---

  • Example policy: First Come, First Serve

    The request that has been received first is serviced first Example:

    This policy is terrible for disk, because:

    The disk head will move back and forth like crazy We must try to minimize head movement to achieve shorter service time !!!

    ---

    ---

    ---

  • Elevator scheduling:

    The disk head sweeps in-and-out (like an elevator) When the disk head is on a cylinder k: Disk will service all requests for that cylinder before moving to the next cylinder

    Graphically:

  ---

  ---

  ---

  ---

• Software level solution: pre-fetching
  • The program can use pre-fetching to reduce processing time.

    ---

  • Pre-fetching: (a.k.a. double buffer)

    The program allocates two buffers to process data from a file: Data is read in a buffer: When buffer is full, program processes the data. And at the same time, more data is read in the other buffer: Rotate buffers when done processing data in buffer:

    ---

    ---

  • Note:

    To write such a program, you will need to use multi-threading and synchronization primitives such as semaphores These are topics for another course....