ECS 15/15AT Lecture Notes

Lecture 9: Auxiliary Storage Devices

When you turn off a computer, everything in main memory is erased. The only ways to save the results of your work are either to print out the results or else to store them on an auxiliary storage device. These devices can store information in a form that can be read automatically by machine, and usually they can retain the information for long periods of time.

There are many forms of auxiliary storage, and we will only consider a very few of them in this class. However, the most common technologies for auxiliary storage involve either magnetic storage media (disks, magnetic tape) and optical storage devices. We will touch on other systems only briefly.

Magnetic Tapes

There are many forms that magnetic media can take. Most of them involve a mylar material in which some ferric oxide particles have been embedded. In the case of magnetic tape, the oxide particles are on one side only (disks are coated on both sides). A magnetic tape is divided into tracks; each one used to store a single bit of a byte of data. Most tapes today are 9-track systems, with 8 tracks used to store data bits and the 9th used to record a parity bit. The parity bit is used to check the accuracy of the other 8 bits. If a tape drive is set to check for odd parity, then the total number of bits turned on (i.e. set to 1) must be odd. If the data byte has four bits set to 1, the parity bit is set to one also, giving a total of 5 (i.e., an odd number) of bits set to one; hence, odd parity. If there was some sort of error in reading the character and a bit was misread (for instance a zero bit was read as a one), then there would be an even number of bits on, and a "parity error" would be generated. (Parity is not a very important factor, since errors of this type are increasingly rare, but it explains why there are nine tracks to magnetic tapes which record 8-bit bytes.)

Magnetic tapes are usually 1/2" wide and from 800 to 2400 feet long. The density at which the data are recorded on a tape is measured in bits-per-inch-per-track, or bpi. Today's magnetic tapes record data usually at 1600 bpi or 6250 bpi. Data are written to magnetic tape in blocks usually consisting of 800 bytes or more. Between each block of data is an inter-record gap, a blank piece of tape about 0.6" in length, to give the tape drive time to get the tape moving up to speed before the next data block is read. Typical storage volume on a 1600 bpi tape would be around 30 MB (30 megabytes) of data.

It takes between 3-5 minutes to read a single 2400 foot reel of tape. Therefore, although data can be transferred at high rates from tape to computer (about 120,000 bits per second), the fact that there are 2400 feet of tape makes it impossible to read it any faster. Therefore, if you wanted to get some data that was stored near the end (inside of the reel) of the tape, you would have to read from the beginning until you reached that point. For this reason, magnetic tapes are used today more for long term storage than for on-line processing of data.

There are many other forms of magnetic tape media. Videocassettes can be used to store computer data, and there are some new forms of video-like technology that can use very small video-type disks to store large volumes of data: 1,500 megabytes (referred to as 1.5 Gigabytes). There are also special purpose cards for personal computers that can convert VHS systems designed for television into data storage devices.

Magnetic Disks

A magnetic disk uses the same basics0 surfaces principles: mylar medium coated with ferric oxide, and read/write heads positioned very close to the surfaces (top and bottom) to read the 0s and ones of each data element.

Each disk must be formatted before it can be used. Formatting a disk consists of dividing the disk into sectors, pie-shaped wedges of data separated by very small empty spaces. Within each sector are a series of tracks, concentric rings of data that can be read one at a time. An important point to remember is that, unlike magnetic tapes, the tracks on a disk can be accessed directly; in other words, one can skip to inner tracks without reading intermediate tacks. For this reason, disks are referred to as direct access, or random access devices.

Figure 1

Disks have evolved rapidly since they became commonplace in the mid 1960s. The first disk systems consisted of disk packs, sets of 6 mylar-coated metal platters, with 10 surfaces read by a set of arms holding the read-heads that moved in unison over each surface of the disk. Figure 1 shows a similar set of four read heads for a two-platter disk system like the ones used in personal computers today. We refer to the set of tracks being read by the multiple read heads as a cylinder of data. It takes more time to move from one track to another than to read all the sectors on one track.

Early disk packs stored somewhere between 20 and 80 MBytes of data on one pack. With improving technologies, however, much greater storage capacity has been achieved. In the mid 1970s, a new technique for storing data on sealed disks (not removable) was invented. The inventor (an IBM researcher) wrote a paper describing his method and predicting that it might be possible to store 30 MB on 30 tracks of data. The concept of 30-30 was quickly translated into "Winchester," for the Winchester 30-30 rifle, and though there never has been a Winchester drive of that exact description, the name stuck. Winchester drives have become smaller and smaller, and they stored more and more data. It is now possible to store over 100 Mbytes of data on Winchester drives using 5 1/2" disks. The increase in storage capacity seems far from ended at this time.

2

Around the same time that Winchester disks were appearing, another invention came into being: the floppy disk, a mylar medium that is not attached to a metal surface and is read by heads that actually touch the upper and lower surface of the disk. Floppy disks are very convenient forms of data storage, since they can be removed, mailed, carried in a briefcase or pocket, and read on many different types of computers. For this reason, they are widely used on most personal computers today. Almost all personal computers come with the operating system on floppy disks, and a new computer with a hard drive (a Winchester drive) must first have the operating system transferred from floppy to Winchester so that it can be booted from the hard drive from then on.

Floppy disks are convenient devices to discuss disk technology; the main differences between floppies and hard drives are speed and storage capacity.

All disk systems are direct access devices, meaning that you can get an single item of data in the time it takes to move the read head to the appropriate track. However, in order to know where to go on a disk, it is important to have some sort of directory that tells the operating system where files are stored within the disk.

A blank disk, newly formatted, consists of a set of tracks, each divided into sectors. The very first sector of the first track (track 0) is usually reserved for the boot record, the program used to load an operating system from other parts of the disk. Not all disks have operating systems, but all disks reserve the first sector for a boot record whether it contains that program or not.

The next few sectors (usually tracks 0 and 1) contain two other types of information. First, they have a list of all sectors available on the disk. A newly formatted disk has a marker for each sector; each time a file is created, the operating system looks at that table (called the file allocation table) and picks the first sectors available, starting from the outside of the disk and moving towards the center. When a file is deleted, the sectors it used to occupy are returned to the list of available sectors, and writing a new file will again start from the outside and work towards the center, allocating as many sectors as required for the new file. This storage strategy means that data tend to be stored more on the outside tracks, but it also means that individual files are not necessarily stored contiguously on disk; the sectors allocated are the first ones available at the time the file was written, and they may be in many different parts of the disk.

Following the file allocation table is another reserved portion of the disk that contains the root directory. In a 360K disk, the number of files that can be indexed in the root directory is approximately 128. If a large number of small files were created, it is possible that entry space in the root directory would be exhausted before the available space on the disk is used up.

Since you already know that DOS files are hierarchical, you can imagine that moving from the root directory to a new subdirectory opens up the door to creating many more files. Subdirectories have their index file records located elsewhere on the disk, but there will be information in the root directory as to how to get to all subdirectories. Although you will probably not be concerned with this point, the speed of accessing root directories is much faster than accessing subdirectory data, so in some cases it is useful to put most often used files in the root directory so that they can be accessed more rapidly.

One other point to bear in mind: if you have used a floppy disk for a long time, it is likely that files will become increasingly fragmented. Copying one file at a time (with the most frequently used copied first) to a new disk will probably improve the disk access time for such files.

Magnetic disks remain the most popular medium of auxiliary storage today, as they have for about two decades. The technology continues to advance: manufacturing improvements make it possible to have finer fragments of ferric oxide to magnetize. New encoding techniques ("phase encoding") promise to vastly increase the storage capacity. Servo mechanical mechanisms offer the possibility of variable speed drives, increasing reliability especially in the inner portions of disks. In short, there is little reason to believe that we will be getting rid of magnetic disks for the next decade or so.

The negative aspect of magnetic disks is that the data stored can be destroyed by very strong magnetic fields and by heat. Therefore, you should be careful (although airplane xray detection units are apparently not strong enough to affect them). Reasonable care is all that is required to give disks systems a long life. The only remaining uncertainty is the obsolescence created by technological advances. For instance, it is very hard to find a system operational today that can read 8" disks, even if the data stored on them is still valid (and it probably is). Whereas magnetic tape drives have remained relatively stable for about twenty years, the disk technology continues to make it likely that the "old" drives of today may be hard to read in another ten years, just as disks written as little as six years ago are hard to read today. It seems likely that the 5 1/2" drive will, in the next five years, become obsolete as the 3 1/2" disk takes its place.

Optical Storage Devices

A second technology that lends itself to storing data in auxiliary form is based on the use of laser beams to create small disruptions in some type of medium. By shining a tightly focused laser beam on a substance that can be melted by the intensity of the beam, small bubbles are created, and, when a lower intensity laser is pointed at the surface, light does not bounce back. By alternating bubbles and smooth, reflecting original surface areas, it is possible to store digital data in a form that can be converted into bits and bytes (also into graphic images and other forms of data including, of course, music).

Optical data disks come in three basic types: read only, non-erasable write-once, and erasable read/write. The read-only disks are currently distributed in the same basic format as Compact Disks (CDs) used to record music. It is fairly expensive to make the original master (costs have dropped to about $1200-2,000 from much higher figures a few years ago), but duplicates are considerably cheaper to make (though commercial CD data disks tend to sell for close to $100 because they have so much valuable data on them). A CD can hold very large amounts of information, probably around 300MBytes, but often they are distributed with far less information because it is hard to put together single collections of data requiring that much storage. A growing industry is springing up of CDs with programs and data (encyclopedias and the like). Since CD data readers for PCs are relatively cheap (under $1,000), they offer a potential for distributing very large amounts of static data relatively cheaply.

Write Once, Read Many (WORM) optical drives are interesting devices. They store large amounts of data (typically 500MBytes to 1 Gigabyte) on a single disk. Once an item is written on a disk, that area is permanently transformed and cannot be reused. This characteristic is helpful in many respects, but it does create fairly complex indexing problems with updates to files (how do you link the old file to the new one?) which opens the door to interesting facets of research for computer scientists. WORM disks are typically larger than CDs (I believe they runaround 10" on average). They cost about $100 each, but the amount of data they store makes them worth it. Many WORM drives are designed to have replaceable disks, so that the total amount you can store on them depends only on how many disks you purchase. An important use of WORM drives is archival storage: keeping data for long periods of time but in a form where it is less likely to be needed for immediate on-line access.

WORM technology has also been converted to a credit card-sized device that does not rotate, instead has a series of rows on which data can be stored. These card are being used to allow patients to carry their own medical records with them (One card holds 2.5 MB of data; more than needed for any patient record), and other applications will probably develop in time.

The final form of optical disk uses a substance that can be melted like the material in WORM and CD drives, but it can also be restored to its original condition by a general heating process. These drives are slower than magnetic disk drives of the same storage capacity, but they are perhaps slightly less expensive. In addition, like all optical storage, they cannot be damaged by magnetic fields (though extreme heat will damage them), and hence they may serve in some specific applications where magnetic disks might have trouble. Sony makes one type of read/write optical disk. But they are quite slow compared to magnetic disks and have never become popular. The new Digital Video Disk will undoubtedly be increasingly important in the years ahead, but at present (Fall, 2000) they are not yet very widespread.

Alternate types of disks are continuing to appear, such as ZIP drives, which compress data so that large volumes of data (>100Mbytes) can be stored and then transported to other systems. New forms of backup tapes are constantly being devised, some of them capable of storing very large volumes of data to back up hard drives. The main points to remember about auxiliary storage devices is their ability to preserve large volumes of data in a form that is difficult to destroy, perhaps difficult to change, and are relatively cheap in a bytes-per-dollar measure. In order for such devices to be successful, they must be standardized, and one hopes that the technology will not make the devices outdated too soon.