Computer storage technology

Memory hierarchy

Memory hierarchy refers to the different types of memory devices and equipment configured into an operational computer system to provide the necessary attributes of storage capacity, speed, access time, and cost to make a cost-effective practical system. The fastest-access memory in any hierarchy is the main memory in the computer. In most computers, random-access memory (RAM) chips are used because of their high speed and low cost. The secondary storage in the hierarchy usually consists of disks. The last, or bottom, level (sometimes called the tertiary level) of storage hierarchy is made up of magnetic tape transports and mass-storage tape systems. Performance is usually measured by two parameters: capacity and access time. (Speed or data rate is a third parameter, but it is not so much a function of the device itself as of the overall memory design.) Capacity refers to the maximum on-line user capacity of a single connectable memory unit. Access time is the time required to obtain the first byte of a randomly located set of data. See Bit

Memory organization

The efficient combination of memory devices from the various hierarchy levels must be integrated with the central processor and input/output equipment, making this the real challenge to successful computer design. The resulting system should operate at the speed of the fastest element, provide the bulk of its capacity at the cost of its least expensive element, and provide sufficiently short access time to retain these attributes in its application environment. Another key ingredient of a successful computer system is an operating system (that is, software) that allows the user to execute jobs on the hardware efficiently. Operating systems are available which achieve this objective reasonably well. See Computer systems architecture

The computer system hardware and the operating system software must work integrally as one resource. In many computer systems, the manufacturer provides a virtual memory system. It gives each programmer automatic access to the total capacity of the memory hierarchy without specifically moving data up and down the hierarchy and to and from the central processing unit (CPU). See Computer programming, Database management system, Operating system, Programming languages

A cache memory is a small, fast buffer located between the processor and the main system memory. Cache memory is used to speed up the flow of instructions and data into the central processing unit from main memory. This cache function is important because the main memory cycle time is typically slower than the central processing unit clocking rates.

Main memory

Random access memory (RAM) chips come in a wide variety of organizations and types. Computer main memories are organized into random addressable words in which the word length is fixed to some power-of-2 bits (for example, 4, 8, 16, 32, or 64 bits). But there are exceptions, such as 12-, 18-, 24-, 48-, and 60-bit word-length machines. Usually RAMs contain NK · 1 (for example, 64 · 1) bits, so the main memory design consists of a stack of chips in parallel with the number of chips corresponding to that machine's word length. There are two basic types of RAMs, static and dynamic. The differences are significant. Dynamic RAMs are those which require their contents to be refreshed periodically. They require supplementary circuits on-chip to do the refreshing and to assure that conflicts do not occur between refreshing and normal read-write operations. Even with those extra circuits, dynamic RAMs still require fewer on-chip components per bit than do static RAMs (which do not require refreshing).

Static RAMs are easier to design, and compete well in applications in which less memory is to be provided, since their higher cost then becomes less important. They are often chosen for minicomputer memory, or especially for microcomputers. Because they require more components per chip, making higher bit densities more difficult to achieve, the introduction of static RAMs of any given density occurs behind that of dynamic versions.

There is another trade-off to be made with semiconductor RAMs in addition to the choice between static and dynamic types, namely that between MOS and bipolar chips. Biopolar devices are faster, but have not yet achieved the higher densities (and hence the lower costs) of MOS. Within each basic technology, MOS and bipolar, there are several methods of constructing devices, and these variations achieve a variety of memory speeds and access times, as well as power consumption and price differences. Within the basic MOS technologies there are several types, such as the n-channel MOS referred to as NMOS and the complementary MOS solid-state structure referred to as CMOS. For bipolar there are several types such as transistor-to-transistor logic (TTL) and the emitter-coupled logic (ECL). See Logic circuits

Secondary memory

High-capacity, slower-speed memory consists of two major functional types: random-access, which has been provided primarily by disk drives, and sequential-access, which has been provided primarily by tape drives. Since tape drives provide removability of the medium from the computer, tape is used for the majority of off-line, archival storage, although some disks are removable also. The on-line random-access disk devices are classed as secondary, and tape-based systems are classed as tertiary.

Conventional magnetic-disk memories consist of units which vary in capacity from the small floppy disks to gigabyte and higher capacity disk drives. The major area of development in disks has been the progressive and even spectacular increases in capacity per drive, particularly in terms of price per byte.

Optical recording is a nonmagnetic disk technology that uses a laser beam to burn pits in the recording medium to represent the bits of information, and a lower-power laser to sense the presence or absence of pits for reading.

CD-ROM (compact disk-read-only memory) and WORM (write once, read many) are special types of optical disks. CD-ROM resembles the related audio compact disk technology in that users of CD-ROM can read only prerecorded data on the disk. A 5-in. (125-mm) CD-ROM can hold 500 to 600 megabytes, which is equivalent to 1400 to 1700 (360-kilobyte) diskettes.

Bubble memories are chips rather than disks, but are different from semiconductor memories in that they are magnetic devices, in which the absence or presence of a magnetic domain is the basis for a binary 1 to 0. The performance characteristics of these devices makes them competitive as small-capacity secondary storage. For portable and other special applications, bubbles have definite advantages such as nonvolatility, low power, and high compactness. Performance capabilities relative to floppy disks are 100 kilobits per second for bubbles versus 200–250 kilobits per second for floppies, and 40 milliseconds average access time for bubbles versus 200– 250 milliseconds for floppies.

Magnetic tape units

In magnetic tape units, the tape maintains physical contact with the fixed head while in motion, allowing high-density recording. The long access times to find user data on the tape are strictly due to the fact that all intervening data have to be searched until the desired data are found. This is not true of rotating disk memories or RAM word-addressable main memories. The primary use of tape storage is for seldom-used data files and as back-up storage for disk data files. Half-inch (12.5-mm) tape has been the industry standard since it was first used commercially in 1953. Half-inch magnetic tape drive transports are reel-to-reel recorders with extremely high tape speeds (up to 200 in. or 5 m per second), and fast start, stop (on the order of 1 millisecond), reverse, and rewind times. Performance and data capacity of magnetic tape have improved by orders of magnitude.

Mass storage systems

With the gradual acceptance of virtual memory and sophisticated operating systems, a significant operational problem arose with computer systems, particularly the large-scale installations. The expense and attendant delays and errors of humans storing, mounting, and demounting tape reels at the command of the operating system began to become a problem. Cartridge storage facilities are designed to alleviate this problem.

Their common attributes are: capacity large enough to accommodate a very large database on-line; access times between those of movable-head disks and tapes; and operability, without human intervention, under the strict control of the operating system. The cartridge storage facility is included within the virtual address range. All such configurations mechanically extract from a bin, mount on some sort of tape transport, and replace in a bin, following reading or writing, a reel or cartridge of magnetic tape.

Cartridge storage systems are hardware devices that need operating system and database software in order to produce a truly integrated, practical hardware-software system. Users require fast access to their files, and thus there is a definite need to queue up (stage) files from the cartridge storage device onto the disks. The database software must function efficiently to make this happen. In general, users base their storage device selection on the file sizes involved and the number of accesses per month. Magnetic tape units are used for very large files accessed seldom or infrequently. Mass-storage devices are for intermediate file sizes and access frequencies. Disk units are used for small files or those which are accessed often.