Mainframes (often colloquially referred to as Big Iron) are computers used mainly by large organizations for critical applications, typically bulk data processing such as census, industry and consumer statistics, ERP, and financial transaction processing.
The term probably originated from the early mainframes, as they were housed in enormous, room-sized metal boxes or frames.  Later the term was used to distinguish high-end commercial machines from less powerful units which were often contained in smaller packages.
Today in practice, the term usually refers to computers compatible with the IBM System/360 line, first introduced in 1965. (IBM System z9 is IBM’s latest incarnation.) Otherwise, systems with similar functionality but not based on the IBM System/360 are referred to as “servers.” However, “server” and “mainframe” are not synonymous (see client-server).
Some non-System/360-compatible systems derived from or compatible with older (pre-Web) server technology may also be considered mainframes. These include the Burroughs large systems, the UNIVAC 1100/2200 series systems, and the pre-System/360 IBM 700/7000 series. Most large-scale computer system architectures were firmly established in the 1960s and most large computers were based on architecture established during that era up until the advent of Web servers in the 1990s. (Interestingly, the first Web server running anywhere outside Switzerland ran on an IBM mainframe at Stanford University as early as 1990. See History of the World Wide Web for details.)
There were several minicomputer operating systems and architectures that arose in the 1970s and 1980s, but minicomputers are generally not considered mainframes. (UNIX arose as a minicomputer operating system; Unix has scaled up over the years to acquire some mainframe characteristics.)
Modern mainframe computers have abilities not so much defined by their single task computational speed (flops or clock rate) as by their redundant internal engineering and resulting high reliability and security, extensive input-output facilities, strict backward compatibility for older software, and high utilization rates to support massive throughput. These machines often run for years without interruption, with repairs and even software and hardware upgrades taking place during normal operation. For example, ENIAC remained in continuous operation from 1947 to 1955. More recently, there are several IBM mainframe installations that have delivered over a decade of continuous business service as of 2007, with upgrades not interrupting service. Mainframes are defined by high availability, one of the main reasons for their longevity, as they are used in applications where downtime would be costly or catastrophic. The term Reliability, Availability and Serviceability (RAS) is a defining characteristic of mainframe computers.In the 1960s, most mainframes had no interactive interface. They accepted sets of punch cards, paper tape, and/or magnetic tape and operated solely in batch mode to support back office functions, such as customer billing. Teletype devices were also common, at least for system operators. By the early 1970s, many mainframes acquired interactive user interfaces and operated as timesharing computers, supporting hundreds or thousands of users simultaneously along with batch processing. Users gained access through specialized terminals or, later, from personal computers equipped with terminal emulation software. Many mainframes supported graphical terminals (and terminal emulation) by the 1980s (if not earlier). Nowadays most mainframes have partially or entirely phased out classic user terminal access in favor of Web user interfaces.Historically mainframes acquired their name in part because of their substantial size and requirements for specialized HVAC and electrical power. Those requirements ended by the mid-1990s, with CMOS mainframe designs replacing the older bipolar technology. In fact, in a major reversal, IBM touts the mainframe’s ability to reduce data center energy costs for power and cooling and reduced physical space requirements compared to server farms.
Characteristics of mainframes:
Nearly all mainframes have the ability to run (or host) multiple operating systems and thereby operate not as a single computer but as a number of virtual machines. In this role, a single mainframe can replace dozens or even hundreds of smaller servers, reducing management and administrative costs while providing greatly improved scalability and reliability.Mainframes can add system capacity nondisruptively and granularly. Modern mainframes, notably the IBM zSeries and System z9 servers, offer three levels of virtualization: logical partitions (LPARs, via the PR/SM facility), virtual machines (via the z/VM operating system), and through its operating systems (notably z/OS with its key-protected address spaces and sophisticated goal-oriented workload scheduling,[clarify] but also Linux and Java). This virtualization is so thorough, so well established, and so reliable that most IBM mainframe customers run no more than two machines: one in their primary data center, and one in their backup data center—fully active, partially active, or on standby—in case there is a catastrophe affecting the first building. All test, development, training, and production workload for all applications and all databases can run on a single machine, except for extremely large demands where the capacity of one machine might be limiting. Such a two mainframe installation can support continuous business service, avoiding both planned and unplanned outages.Mainframes are designed to handle very high volume input and output (I/O) and emphasize throughput computing. Since the mid-1960’s, mainframe designs have included several subsidiary computers (called channels or peripheral processors) which manage the I/O devices, leaving the CPU free to deal only with high-speed memory. It is common in mainframe shops to deal with massive databases and files. Giga-record or tera-record files are not unusual. Compared to a typical PC, mainframes commonly have hundreds to thousands of times as much data storage online, and can access it much faster.
Mainframe return on investment (ROI), like any other computing platform, is dependent on its ability to scale, support mixed workloads, reduce labor costs, deliver uninterrupted service for critical business applications, and several other risk-adjusted cost factors. Some argue that the modern mainframe is not cost-effective. Hewlett-Packard and Dell unsurprisingly take that view at least at times, and so do a few independent analysts. Sun Microsystems used to take that view but, beginning in mid-2007, started promoting its new partnership with IBM, including probable support for the company’s OpenSolaris operating system running on IBM mainframes. The general consensus (held by Gartner and other independent analysts) is that the modern mainframe often has unique value and superior cost-effectiveness, especially for large scale enterprise computing. In fact, Hewlett-Packard also continues to manufacture its own mainframe (arguably), the NonStop system originally created by Tandem. Logical partitioning is now found in many UNIX-based servers, and many vendors are promoting virtualization technologies, in many ways validating the mainframe’s design accomplishments.
Mainframes also have unique execution integrity characteristics for fault tolerant computing. System z9 servers execute each instruction twice, compare results, and shift workloads “in flight” to functioning processors, including spares, without any impact to applications or users. This feature, also found in HP’s NonStop systems, is known as lock-stepping, because both processors take their “steps” (i.e. instructions) together. Not all applications absolutely need the assured integrity that these systems provide, but many do, such as financial transaction processing.
Despite these differences, the IBM mainframe, in particular, is still a general purpose business computer in terms of its support for a wide variety of popular operating systems, middleware, and applications.
Several manufacturers produced mainframe computers from the late 1950s through the 1970s. The group of manufacturers was first known as “IBM and the Seven Dwarfs“: IBM, Burroughs, UNIVAC, NCR, Control Data, Honeywell, General Electric and RCA. Later, shrinking, it was referred to as IBM and the BUNCH. IBM’s dominance grew out of their 700/7000 series and, later, the development of the 360 series mainframes. The latter architecture has continued to evolve into their current zSeries/z9 mainframes which, along with the then Burroughs and now Unisys MCP-based mainframes, are among the few mainframe architectures still extant that can trace their roots to this early period. That said, while they can still run 24-bit System/360 code, the 64-bit zSeries and System z9 CMOS servers have nothing physically in common with the older systems. Notable manufacturers outside the USA were Siemens and Telefunken in Germany, ICL in the United Kingdom, and Fujitsu, Hitachi, Oki, and NEC in Japan. The Soviet Union and Warsaw Pact countries manufactured close copies of IBM mainframes during the Cold War; the Strela is an example of an independently designed Soviet computer.Shrinking demand and tough competition caused a shakeout in the market in the early 1980s — RCA sold out to UNIVAC and GE also left; Honeywell was bought out by Bull; UNIVAC became a division of Sperry, which later merged with Burroughs to form Unisys Corporation in 1986. In 1991, AT&T briefly owned NCR. During the same period, companies found that servers based on microcomputer designs could be deployed at a fraction of the acquisition price and offer local users much greater control over their own systems given the IT policies and practices at that time. Terminals used for interacting with mainframe systems were gradually replaced by personal computers. Consequently, demand plummeted and new mainframe installations were restricted mainly to financial services and government. In the early 1990s, there was a consensus among industry analysts that the mainframe was a dying market as mainframe platforms were increasingly replaced by personal computer networks.That trend started to turn around in the late 1990s as corporations found new uses for their existing mainframes and as the price of data networking collapsed in most parts of the world. The growth of e-business also dramatically increased the number of back-end transactions processed by mainframe software as well as the size and throughput of databases. Another factor currently increasing mainframe use is the development of the Linux operating system, which can run on many mainframe systems, typically in virtual machines. Linux allows users to take advantage of open source software combined with mainframe hardware RAS. Rapid expansion and development in emerging markets, particularly China, is also spurring major mainframe investments to solve exceptionally difficult computing problems, e.g. providing unified, extremely high volume online transaction processing databases for 1 billion consumers across multiple industries (banking, insurance, credit reporting, government services, etc.)
Mainframes vs. supercomputers:
The distinction between supercomputers and mainframes is not a hard and fast one, but supercomputers generally focus on problems which are limited by calculation speed while mainframes focus on problems which are limited by input/output and reliability (“throughput computing”) and on solving multiple business problems concurrently (mixed workload). The differences and similarities include:
- Both types of systems offer parallel processing. Supercomputers typically expose it to the programmer in complex manners, while mainframes typically use it to run multiple tasks. One result of this difference is that adding processors to a mainframe often speeds up the entire workload transparently.
- Supercomputers are optimized for complicated computations that take place largely in memory, while mainframes are optimized for comparatively simple computations involving huge amounts of external data. For example, weather forecasting is suited to supercomputers, and insurance business or payroll processing applications are more suited to mainframes.
- Supercomputers are often purpose-built for one or a very few specific institutional tasks (e.g. simulation and modeling). Mainframes typically handle a wider variety of tasks (e.g. data processing, warehousing). Consequently, most supercomputers can be one-off designs, whereas mainframes typically form part of a manufacturer’s standard model lineup.
- Mainframes tend to have numerous ancillary service processors assisting their main central processors (for cryptographic support, I/O handling, monitoring, memory handling, etc.) so that the actual “processor count” is much higher than would otherwise be obvious. Supercomputer design tends not to include as many service processors since they don’t appreciably add to raw number-crunching power.
There has been some blurring of the term “mainframe,” with some PC and server vendors referring to their systems as “mainframes” or “mainframe-like.” This is not widely accepted and the market generally recognizes that mainframes are genuinely and demonstrably different.
Speed and performance:
The CPU speed of mainframes has historically been measured in millions of instructions per second (MIPS). MIPS have been used as an easy comparative rating of the speed and capacity of mainframes. The smallest System z9 IBM mainframes today run at about 26 MIPS and the largest about 17,801 MIPS. IBM’s Parallel Sysplex technology can join up to 32 of these systems, making them behave like a single, logical computing facility of as much as about 569,632 MIPS.The MIPS measurement has long been known to be misleading and has often been parodied as “Meaningless Indicator of Processor Speed.” The complex CPU architectures of modern mainframes have reduced the relevance of MIPS ratings to the actual number of instructions executed. Likewise, the modern “balanced performance” system designs focus both on CPU power and on I/O capacity, and virtualization capabilities make comparative measurements even more difficult. See benchmark (computing) for a brief discussion of the difficulties in benchmarking such systems. IBM has long published a set of LSPR (Large System Performance Reference) ratio tables for mainframes that take into account different types of workloads and are a more representative measurement. However, these comparisons are not available for non-IBM systems. It takes a fair amount of work (and maybe guesswork) for users to determine what type of workload they have and then apply only the LSPR values most relevant to them.To give some idea of real world experience, it is typical for a single mainframe CPU to execute the equivalent of 50, 100, or even more distributed processors’ worth of business activity, depending on the workloads. Merely counting processors to compare server platforms is extremely perilous.