CS139-lecture-20211202

Mass storage #

Magnetic disks #

image_2021-12-02-13-45-05 image_2021-12-02-13-48-46 image_2021-12-02-13-50-17

  • we want to keep the same rate of data moving under the head.

image_2021-12-02-13-53-52 image_2021-12-02-13-55-44

  • a track is 2D but the entire disk is 3D

image_2021-12-02-13-56-38 image_2021-12-02-13-56-43

Watch: Inside of a hard disk drive

image_2021-12-02-14-01-33

Disk performance #

image_2021-12-02-14-01-42 image_2021-12-02-14-01-48 image_2021-12-02-14-05-32

  • seek and rotate times dominate the total latency
  • hard disks are at the milisecond level

image_2021-12-02-14-07-51

Disk history #

image_2021-12-02-14-08-37 image_2021-12-02-14-08-42

Solid state disks #

image_2021-12-02-14-08-51

  • SSDs are generally faster because they do not have any moving parts, thus no seek and rotate times, this greatly speeds up the total latency
  • SSDs are at the microsecond level

Magnetic tape #

image_2021-12-02-14-11-35

  • magnetic tape remains in use today as a tertiary storage
  • ideal for backup/logs

Disks vs memory #

image_2021-12-02-14-11-49

Disk scheduling #

The one component of disk latency that we have the ability to optimize is the seek time. If we minimize the seek distance we can minimize the seek time.

image_2021-12-02-14-24-25 image_2021-12-02-14-24-34 image_2021-12-02-14-24-47 image_2021-12-02-14-25-01 image_2021-12-02-14-26-21

  • needs to know the entire sequence in advance

image_2021-12-02-14-27-56 image_2021-12-02-14-28-02 image_2021-12-02-14-34-57 image_2021-12-02-14-35-05

  • remember that the “giant leap” here is still a physical movement

image_2021-12-02-14-39-34 image_2021-12-02-14-39-41

Selecting a disk scheduling algorithm #

image_2021-12-02-14-40-33

RAID #

image_2021-12-02-14-44-36 image_2021-12-02-14-44-39 image_2021-12-02-14-44-44 image_2021-12-02-14-44-53