Scribd
Upload
39 views

Lecture 20: Hard Disk Internals: Mythili Vutukuru IIT Bombay

Hard disk internals consist of one or more spinning platters divided into tracks and 512-byte sectors. When accessing a sector, the disk head moves to the correct track (seek time) and waits for the desired sector to rotate under the head (rotational latency). Disk scheduling algorithms like SSTF, Elevator, and SPTF reorder requests to minimize seek times and rotational latencies. Error detection/correction bits store data reliably on disk despite bit flips, with disk controllers handling some errors and RAID providing high reliability through replication.

Uploaded by

Ravi Chandra Reddy
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
39 views

Lecture 20: Hard Disk Internals: Mythili Vutukuru IIT Bombay

Hard disk internals consist of one or more spinning platters divided into tracks and 512-byte sectors. When accessing a sector, the disk head moves to the correct track (seek time) and waits for the desired sector to rotate under the head (rotational latency). Disk scheduling algorithms like SSTF, Elevator, and SPTF reorder requests to minimize seek times and rotational latencies. Error detection/correction bits store data reliably on disk despite bit flips, with disk controllers handling some errors and RAID providing high reliability through replication.

Uploaded by

Ravi Chandra Reddy
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 9

Lecture 20: Hard disk internals

Mythili Vutukuru
IIT Bombay
Hard disk
• Interface: a set of 512-byte blocks (sectors),
that can be read or written atomically
• Internals: one or more platters, connected by
a spindle, spinning at ~10K RPM (rotations per
minute)
• Each platter has a disk head and arm
• A platter is divided into multiple tracks, and
each track into 512-byte sectors

2
What happens when accessing a
sector?
• Suppose disk head at 30, need to access 11
• Seek to the correct track, wait for disk to rotate

3
Time taken for I/O operation
• Time taken to read/write a block consists of
– Seek time to get to the right track (few ms)
– Rotational latency for disk to spin to correct sector on
the track (few ms)
– Transfer time to read sector (few tens microsec)
• Given high seek and rotational latency, usually
rate of sequential access > rate of random access

4
Disk Scheduling
• Requests to disk are not served in FIFO, they
are reordered with other pending requests
• Why? In order to read blocks in sequence as
far as possible, to minimize seek time and
rotational delay
• Who does scheduling? OS does not know
internal geometry of disk, so scheduling done
mostly by disk controller

5
Shortest Seek Time First (SSTF)
• Access block that we can
seek to fastest
– Go to 21 (one track away)
before 2 (two tracks away)
• Problem: starvation
(some requests that are
far from current position
of head may never get
served)

6
Elevator/SCAN algorithm
• Disk head does one sweep over tracks and
serves requests that fall on the path
• Elevator/SCAN: sweep outer to inner, then
inner to outer
• C-SCAN: sweep only one direction (say, outer
to inner) and circle back, start again
– Why? Sweeping back and forth favors middle
tracks more
• F-SCAN: freeze queue while scanning
– Why? Avoid starving far away requests
7
Shortest Positioning Time First (SPTF)
• Considers both seek time
and rotational latency
– Better to serve 8 before 16,
even though seek time is
higher
– Why? 16 incurs a much
higher rotational latency

8
Data storage on disk
• Bits stored on disk with some error
detection/correction bits
– Correct random bit flips
– Detect corruption of data
• Disk controller or OS can handle some errors
(e.g., blacklisting certain sectors)
• If errors cannot be masked, user perceives hard
disk failures
• Technologies such as RAID (Redundant Array of
Inexpensive Disks) provide high reliability and
performance by replicating across multiple disks.
9

You might also like