views updated May 29 2018


Prior to the advent of the personal computer or PC, the minicomputer, and the microcomputer, the term "computer" simply referred to mainframes. What differentiates the modern mainframe from these other classes of computers is the scope of the processing taking place. The typical mainframe today serves tens of thousands of users processing thousands of transactions every second while maintaining centralized terabyte -size databases. Even the mighty supercomputer, although unquestionably faster doing one thing at a time, is not up to this task.

Surprisingly, the processors used in these machines are not much faster than those found in PCs. The architecture of the mainframe, however, provides that the processors can work together in parallel and focus primarily on the actual processing of data and instructions. They are relieved of the time-consuming duties of controlling input and output. Mundane tasks such as reading data from disk, handling transmissions to user terminals, and even reading and writing into main memory, are handled by sub-systems that may well be as powerful as the main processor.

Although a PC is designed to provide very fast processing to a single user, the mainframe must be able to control many tasks being run by many users simultaneously. Thus, the mainframe is differentiated from other computing systems in the areas of data bandwidth , organization, reliability, and control.

Evolution of Mainframes

In the early 1960s, companies such as Burroughs, IBM, RCA, NCR, and Sperry Rand manufactured mainframes. Since the 1970s, the only mainframes in use are the System/390, made by IBM, or clones made by Hitachi and Fujitsu. This provides a high degree of hardware compatibility both within and across manufacturer lines. The System/390 evolved from the System/360, which was initially introduced by IBM in 1965.

Over time, the physical size and cost of mainframes have been reduced dramatically. What once cost millions and filled a large data center can now literally fit in a single 48.7-centimeter (19-inch) wide cabinet for a few hundred thousand dollars. Power consumption and heat dissipation have also been reduced. In the 1970s and 1980s, mainframes gave off so much heat they had to be cooled with chilled water.

The fifth generation machines on the market now use Complimentary Metal Oxide Semiconductor, or CMOS, technology and require no special cooling. They can provide up to 12 parallel 650 MHz processors, 32 GB memory, and several terabytes of disk storage. Compare that to a typical PC with a single 500 MHz processor, 64 MB of memory, and 10 GB of disk storage!

Mainframe vs. Personal Computer

Another contrast to the PC is the type of jobs the mainframes run. Transaction processing jobs run constantly in real-time and must be available more than 99.99 percent of the time. The reboots and lock-ups common with PCs are simply not acceptable. Thousands of individual users can log in simultaneously from a variety of sources such as computer terminals, ATM machines, or Internet web sites, and complete a single transaction.

Time-sharing jobs can be started when needed from a computer terminal by authorized users who then use the mainframe as their own big PC. Finally, batch jobs are started automatically by the system at regular times according to a strict predetermined schedule. Batch jobs are used to do the periodic processing required on the data being received from transaction and time-sharing jobs. Closing the accounting books at month-end or copying disk files to tape for backup are examples of batch type processing.

Mainframe Components

Unlike the single box used by the PC, the mainframe has many components. Typically, they include the following.

Operating System.

Although many would argue that the operating system (OS) is software and therefore not a component, it is in fact the most important and complex component of the mainframe. The OS used in the largest of mainframes, including clones, is IBM's OS/390, running Multiple Virtual Systems (MVS). The OS/390 also runs Virtual Memory (VM), VSE, and Unix. MVS has proven itself to be the only OS capable of handling the multiple processing requirements of today's largest businesses.

A key feature of MVS is Job Control Language (JCL), which enables the operators to automate all jobs, scheduling their running times, and handling exceptions that occur. Within MVS, a software sub-system called Customer Information Control System (CICS) enables the concurrent processing of transactions by thousands of users. There are many other subsystems such as IMS and DB/2 for database, Time-Sharing Option (TSO) for user jobs, VTAM for telecommunications, and RACF for security. The OS runs the overall system and provides the environment in which application programs can do the actual jobs users want to accomplish. The vast majority of mainframe application programs are written in COBOL (Common Business Oriented Language).

Central Processing Unit.

Similar to the processor in a PC, the central processing unit (CPU) decodes instructions, performs calculations, and issues instructions to other components. Unlike the PC, however, the mainframe will typically have many CPUs connected in parallel. It is the job of the OS to dispatch jobs to parallel processors in an efficient manner. To date, this has proven to be a formidable task. Although once a separate component of the mainframe, main storagealso known as high speed memory or random access memory (RAM) is now typically packaged within the CPU. This is done to increase the speed of data transfer.


The job of the channel is to connect the CPU and main storage to other components of the mainframe configuration. The channel ensures that data are moved in an orderly fashion and verifies the integrity of the data. The cables used for these channels once consisted of dozens of small coaxial cables that created quite a congested area under the computer room floor. Fiber optic cable has resolved this problem while dramatically increasing data transfer speeds.

Disk Storage Subsystems

Disk storage subsystems provide for the long term storage and quick retrieval of large amounts of data. They use a magnetically coated disk spinning at high speed. The latest in this technology incorporates RAID, a random array of inexpensive disks. RAID spreads data across many small disk drives similar to those used in a PC, and uses a method called parity to recreate data should one drive fail. The bad drive can then be replaced without turning off the system or losing any data.

Mass Storage Devices.

Over the years, many schemes have been developed to increase the capacity of data storage. Recent advances in disk storage technology have rendered obsolete many of these devices, such as magnetic strip and optical disk storage. One device that remains in common use is the tape drive. Tape is used today primarily as a means to back up data. These tapes can then be stored somewhere else to be used if a major disaster destroys the data center. In situations where truly massive amounts of data must be archived for occasional use, tapes are organized into libraries and accessed using automated equipment to find and load the tape.

Communications Controllers.

To gain access to a mainframe computer, a user will typically use a dumb terminal or a PC programmed to act as such. The job of keeping track of which user is at which terminal and transferring data is performed by the communications controllers. The connection may be established over a local direct connection, a local area network (LAN) , or a wide area network (WAN) .

Line Printers.

Although many commentators have said that computers will bring about a paperless society, there is still a tremendous amount of data output in printed form. Printers are normally attached via a print controller, although the largest and fastest can connect directly to the channel. The two major categories of printers are impact and laser. Impact printers are used for multiple carbon copies. Laser printers are the workhorse of the industry and produce the majority of output from mainframes. They not only print text, but also graphics and forms at the same time.


The role of the mainframe has gradually changed from that of a data processor to that of a server, with the processing being done on the user's PC. It has also been modified to interface to the Internet through the addition of TCP/IP protocols, Unix, and Java programming, to enable businesses to connect to their customers over that network. Once the only form of business computer available, the mainframe has survived the PC revolution and maintained an important function in commercial computing.

see also Central Processing Unit; Generations, Computers; Generations, Languages; Memory; Operating Systems.

Ken Doerbecker and Bob Patterson


Williams, Michael R. A History of Computing Technology. Englewood Cliffs, NJ: Prentice Hall, 1985.

Internet Resources

Amdahl Corporation. Millenium Global Servers. <>

Hitachi Data Systems. Guide Your IT Infrastructure into the Next Millennium. <>

IBM Corporation. OS/390 Overview. <>

IBM Corporation. S/390 Parallel Enterprise Server and OS/390 Reference Guide. <>


views updated May 23 2018

main·frame / ˈmānˌfrām/ • n. 1. a large high-speed computer, esp. one supporting numerous workstations or peripherals.2. the central processing unit and primary memory of a computer.


views updated May 18 2018

1. Generally, the combination of central processor and primary memory of a computer system. The term excludes the I/O, backing store, etc., and is sometimes used synonymously with central processor.

2. Any large computer system.