Magnetic Disk

A magnetic disk consists of one or more flat, circular platters made of glass, metal, or plastic, and coated with a magnetic substance. Particles within a small area of the magnetic substance can be polarized magnetically in one of two directions with an electromagnet; an electromagnet can also detect the direction of polarization previously recorded. Thus, magnetic polarization can be used to distinguish 1s and 0s. Electromagnetic read/write heads are used for this purpose.

A drive motor rotates the disk platter(s) about its central axis. On most drives, the motor rotates the disk at a fixed speed. An arm has the read/write head mounted at the end. The arm makes it possible for the head to move radially in and out across the surface of the disk. A head motor controls precisely the position of the arm on the disk.

Most hard disk drives contain several platters, all mounted on the same axis, with heads on each surface of each platter. The heads move in tandem, so they are positioned over the same point on each surface. Except for the top and bottom, each arm contains two read/write heads, which service the surfaces of two adjoining platters.

With the head in a particular position, it traces out a circle on the disk surface as the disk rotates; this circle is known as a track. Since the heads on each surface all line up, the set of tracks for all the surfaces form a cylinder. Each track contains one or more blocks of data. On most disks the surface of the disk platter is divided into equally sized pie shape segments, known as sectors, although the disks on some large computers divide up the track differently. Each sector on a single track contains one block of data, typically 512 bytes, which represents the smallest unit that can be independently read or written. Figure 10.2 shows the layout of a hard disk.

If you assume that the number of bytes in a sector is the same anywhere on the disk, then you can see from the layout that the bits on the disk are more closely packed on the inner tracks than they are on the outer tracks. Regardless of the track, the same angle is swept out when a sector is accessed; thus, the transfer time is kept constant with the motor rotating at a fixed speed. This technique is called CAV, for constant angular velocity. CAV has the advantage of simplicity and fast access.

It is possible to increase the capacity of the disk by utilizing the space at the outer tracks to pack more bits onto the disk. But this would result in a different number of bytes per sector or a different number of sectors per track depending on which track is being accessed. This would make it more difficult to locate the required sector. Notice, too, that with a constant speed motor, the time to move the head over a pie-shaped sector at the edge is the same as that near the center. If there were more bits packed into the outer tracks, the data would be transferred faster at the edge than at the center. Since the disk controller is designed to expect data at a constant speed, it would seem to be necessary to design the motor so that it would slow down when the head was accessing the outer tracks to keep the data transfer speed constant. In this case, the motor speed would be adjusted such that the speed along the track would be constant regardless of the position of the head. This approach is called CLV, for constant linear velocity. The capacity of a CLV disk with the same diameter and bit density is approximately double that of an equivalent CAV disk. Although CLV technology is commonly used with CDs and DVDs, the design makes it more difficult to access individual blocks of data rapidly, so it is rarely used for hard disks.

As a compromise, modern disk drives divide the disk into a number of zones, typically sixteen. This approach is shown in Figure 10.3. The cylinders in different zones have a different number of sectors but the number of sectors within a particular zone is constant.Obviously, the largest number of sectors will be in the zone containing the outermost cylinders, Instead of adjusting the motor speed, the disk controller buffers the data rate so that the data rate to the I/O interface is constant, despite the variable data rate between the controller and the disk. Different vendors call this technique multiple zone recording, zone bit recording (ZBR), or zone-CAV recording (Z-CAV).

The platter on a hard disk drive is made of a rigid material and is precisely mounted. The heads on a hard disk do not touch the surface; rather, they ride on a bed of air a few millionths of an inch above the surface. The location of the heads radially is tightly controlled. This precision allows the disk to rotate at high speed and also allows the designers to locate the tracks very close together. The result is a disk that can store large amounts of data and that retrieves data quickly. A typical hard disk rotates at 5400 revolutions per minute (rpm), 7200 rpm, or even 10,800 rpm.

A photograph of a hard disk assembly showing a disk platter, arm, and read/write head is shown in Figure 10.4. This particular hard disk drive contains three platters and six heads. Only the topmost platter and head are fully visible. The entire assembly is sealed to prevent dirt particles from wedging between the heads and the disk platter, since this situation could easily destroy the drive. Even a particle of cigarette smoke is much larger than the space between the head and the disk. When the disk is stationary, the head rests in a parked position on the edge of the drive. The head has an aerodynamic design, which causes it to rise on a cushion of air when the disk platter spins.

Figure 10.5 shows the operation required to locate an individual block of data. First, the arm moves the head from its present track until it is over the desired track. The time that is required to move from one track to another is known as the seek time. Since the distance between the two tracks is obviously a factor, the average seek time is used as a specification for the disk. Once the head is located over the desired track, the read/write operation must wait for the disk to rotate to the beginning of the correct sector. The time for this to occur is known as the rotational latency time, or sometimes as rotational delay or simply latency time. The latency time is obviously variable, depending on the position of the disk. As a best case, the head is just about to enter the sector, and the rotational latency time is 0.

At the opposite extreme, the head has just passed the beginning of the sector, and a full rotation is required to reach the beginning of the sector. This time can be calculated from the rotational speed of the disk. Both situations are equally probable. On average, the disk will have to rotate half way to reach the desired block. Thus, the average latency time can be calculated from the rotational speed of the disk as

Once the sector is reached, the transfer of data can begin. Since the disk is rotating at a fixed speed, the time required to transfer the block, known as transfer time, is defined by 

Since the total time required to access a disk block is approximately the sum of these three numbers, a typical disk access might require 20 to 25 msec. To put these speeds in perspective, consider that the typical modern computer can execute an instruction in less than 1 nanosecond. Thus, the CPU is capable of executing millions of instructions in the time required for a single disk access. This should make it very clear to you that disk I/O is a major bottleneck in processing and also that it is desirable to find other work that the CPU can be doing while a program is waiting for disk I/O to take place.

An expansion of part of a track to show a single data block is shown in Figure 10.7. The block consists of a header, 512 bytes of data, and a footer. An interblock gap separates the block from neighboring blocks. Figure 10.8 shows the layout of the header for a Windows-based disk. The track positions, blocks, and headers must be established before the disk can be used. The process to do this is known as formatting the disk. Since the header identifier must be a unique pattern of 1s and 0s, the data being stored must be checked by the disk controller to assure that the data pattern does not accidentally match the header identifier. If it does, the pattern stored on the disk is modified in a known way.

The entire track is laid down as a serial stream of bits. During write and read operations, the bytes must be deconstructed into bits and reconstructed.

Because the transfer speed of the disk is not the same as that required to transfer the block to memory, buffering is provided in the disk controller. The buffer is a first-in, first-out buffer, which receives data at one speed and releases it as required at the other speed. Buffer memory also makes it possible to read a group of blocks in advance so that requests for subsequent blocks can be transferred immediately, without waiting for the disk. Most modern disks provide substantial buffers for this purpose.

The disk controller identifies the disk block to be located, moves the head to the correct track, then reads the track data until it encounters the header for the correct block. Assuming that it is performing a read, it then transfers the data from the disk to a buffer. From the buffer, the data is transferred to conventional memory using DMA. Once the DMA transfer is complete, the disk controller notifies the CPU with a completion interrupt.