Introduction to Magnetic Disks

Material Stolen from Dr. Kent Foster's CSCI 325 Course Notes

Construction
One or more disks are mounted on a spindle. The disks are sometimes called "platters". The spindle is
connected to shaft of a drive motor. The drive motor rotates continuously at a constant speed
(measured in RPMs). The disk material is non-magnetizeable (using aluminum and/or glass), but the
disk surfaces are coated with a very thin magnetizable coating.
Read/write recording head (one or more) for each disk. The R/W heads
are attached to an "access arm". The access arms are attached to a "boom".
The boom is connected to a "step" motor (also called a "servo-motor").
The step motor moves the boom and R/W heads to a track position and
stops there before a read or write operation can occur.

On a hard drive, the read/write heads float on a cushion of air over the
surfaces of the disks. The smaller the air gap, the closer the read/write
head is to surface, the smaller the magnetized spot. So, the smaller the air
gap is made, the more dense the recording of bits will be. Contaminants
on the surface are larger than the air gap:
air gap
.00003 (30 millionths of an inch)
human hair
.003 (3 thousandths of an inch)
paper
.003
smoke particle .0015
dust particle .0003
finger prints .0006
If a contaminant "crashes" into the R/W head, then as with a car going over a large bump, the R/W
head is forced up and then the access arm will slam it down onto the recording surface and gouge the
magnetizable material off the surface, called a head crash. This makes the disk unusable. This was a
significant problem when disk surfaces were exposed to the ambient air which is filled with
particulates.

Hence, newer drives have the disks, spindle, R/W heads, access arms and boom enclosed in a sealed
case that will not allow contamination from the outside environment. This technology was developed
by IBM and was known as Winchester disk technology in the past, but the term has since been
dropped since all hard disk drives now are constructed this way.

Sectors
Each track is divided into a whole number of sectors. Sector size is fixed for the disk. All PC's use
sectored track organization with 512 byte sectors. Some other sector sizes that have been used by
various manufacturers are 256 and 1024 bytes.

A single disk operation = a read or write operation on one track. The minimum amount of data
transferred between the computer and the drive is 1 sector. The max amount transferred in a single
disk operation is 1 whole track of sectors. Files are stored in a whole number of sectors. In other
words, a file that needs 2.5 sectors of space consumes 3 whole sectors.

For most disk drives, all of the tracks have the same number of sectors. Other disk drives have groups
of tracks with different numbers of sectors. If not all sectors have the same size, then inner disk tracks
will have fewer sectors than outer tracks since the inner tracks have a smaller circumference than
outer tracks. Some hard drives to have 40+ sectors on inner tracks and 50+ sectors on the outer tracks.
If all tracks have same number of sectors, then the linear recording density of the outer track is less
than the inner tracks. But, the tracks hold the same amount of data and so we will not use the linear
recording density of a track as a useful measure. Rather, we use the track capacity measured in
bytes/track (B/trk) as the track measure (as opposed to bytes/inch for drives that have different # of
sectors.)
For example, the old 1.44 MB 3.5" diskette with 18 sectors per track and 80 tracks per surface has
track capacity:
cap =
=
=
=

512 bytes/sector * 18 sectors/track * 80 tracks/surface * 2 surfaces

1,474,560 bytes
1440 KB
1.44 MB

On most disk drives, each sector begins with a small header that contains a gap and address mark.
Most manufacturers also end each sector with an error correcting code. (For the students who have
taken Networking course, think "check bits".)
Exercise #1:

Assuming:

File Size = 500,000 bytes

Sectors per Track = 63
Sector Size = 512 bytes

Determine:
# sectors for the file =
min # tracks for the file =

Access Time
There are 3 delay times involved in accessing and transferring data between the computer and the disk
drive.
- Seek time delay.
- Rotational time delay.
- Transfer time delay.

It is **VERY** important to note that all these calculations are omitting the time necessary for the
CPU to transmit data along the system bus to/from the disk drive controller.

Seek Time
The seek time for a disk operation is the time required to move the read/write head from the current
track position to the target track position.
What causes the delay?
- The R/W head is stopped over the current track.
- The step motor must move the stopped R/W head across the surface to the target track
- The R/W head must become stationary over the target track
Characteristic seek time is determined by
the "head movement profile" described by
the graph. There are three time components
in this movement profile:

1. start up time - the time for the R/W

head to start moving from rest until it
is traveling at maximum linear speed
2. travel time at max speed - the time
the head travels at maximum linear speed until it begins to slow down
3. stop down time - the time required to bring the R/W head from its max. speed to a stationary
position over the target track

For short track to track head moves, the drive never reaches max head movement speed. Just have
some start-up time and stop-down time.

Long track to track head moves are called "Full stroke seek". For example, movement from the outer
track to the inner track, or vice versa. Startup and stop down times are a small percentage of the total
full stroke seek Travel time mainly spent passing over other tracks at maximum head movement
speed.
Most manufacturers only give average read and write seek time. For example, WD Cavier 10.1GB is
9.5 ms.

There is no way for the programmer to know how much seek time is required because the position of
the R/W head is unknown before the disk operation. The seek time is Non-deterministic (cannot be
determined absolutely).

So we must measure seek time "statistically", that is, using statistical measures of the seek time value:
- Minimum seek time,
- Average seek time
- the "expected" seek time, denoted E(seek)
- Maximum seek time, Max(seek)
Typical values:
- Min seek is between 1 ms up to < 10 ms

- Avg seek 10 ms to 20 ms
- Max seek from < 20 ms to > 30 ms

Rotational Latency
The rotational latency for a disk operation is the time required to rotate the disk to a position where
the start of target data is under the read/write head. The rotational latency is based on rotation speed.
Latency is non-deterministic (it cannot be determined absolutely for any given disk operation, in
general), and so must be treated "statistically", that is, using statistical measures of the latency:
- Max latency = time of a full revolution = sec / rev
- Average latency = time of revolution = of max latency

Disk rotation speed is measured in Revolutions Per Minute (RPMs). To determine rotational latency,
we need to know the time for one rotation - in other words, minutes per rotation. Hence, latency is the
inverse of speed.

Time of revolution can be computed as the inverse of revolutions per time. So, time per revolution = 1
/ ( rev / time ).
For example, assume we have a 3600RPM drive,
Max Latency =

1
----------------rev / time

1
----------------3600rev / 60 secs

60 secs
----------------3600 Rev

1 sec
----------------60 Rev

= 16.67ms per revolution

1000ms
---------60 rev

16.67ms
--------1 rev

So, the max latency = 16.67ms and the average latency = 8.33ms.

Transfer Time
The transfer time for a disk operation is the time required to transfer the data from (or to) the disk
surface to (or from) the computer once the start of the data is under the R/W head.

Transfer time is based on:

1 - speed of rotation
2 - density of data on the track
3 - amount of data to be transferred

The rotational speed of the drive and the track capacity, or track density, can be combined into a
single value which is the transfer rate. Remembering that one track = 1 revolution, we find ...
Transfer Rate = track capacity / rotational speed
and so, the time to transfer a set of data is
Transfer Time = Transfer Rate * Amount to Transfer
For example, the HP9154A:

Rotation Speed = 3000RPM

Sector Size = 256 Bytes
Track Capacity = 28 sectors

So, transfer rate = 7 KB / 20ms

= .35 KB / ms
= 3.5KB / sec

--> 20ms/rev

--> 7KB / track

Note that the manufacturer's transfer rate will vary because of the block headers briefly mentioned
above.

Example
Let's look at the manufacturer's spec sheet for the Seagate Barracuda 7200.10 Serial ATA, model
number ST3750640AS.
Capacity
750 GB
Discs
4
Sectors per Track
63
Bytes per Sector
512
Speed
7200 RPM
Avg Latency
4.16 ms
Track-to-Track Seek Time <0.8 ms for read; <1.0 ms per write
Average Seek, Read
< 8.5 ms
Average Seek, Write
< 10.0 ms
seagate spec sheet, see page 10

Exercise #2:

1 - We computed an Average Latency for a 3600 RPM drive to be 8.33ms. So, how can Seagate say
their 7200RPM device has a 4.16ms Average Latency?
2 - How many tracks are on one each side of their platters?