In the strictest sense; the term “mainframe computer” is generally regarded as meaning those computers that are compatible with the IBM System/360 series first introduced in 1965. Other very high-end computers that are not compatible with the IBM System/360 series are usually referred to as “servers”.

Now however; the term “mainframe computer” is typically used to refer to that group of high-end self-contained computers which incorporate copious numbers of built in “hot swap” capable redundant systems to provide added robustness as standard fare, as opposed to less well endowed server class computer systems.

The latter group includes collectives of grouped and associated computers known as server farms that may provide additional resilience through duplicated systems but on the whole do not provide the critical total internal system component redundancy of the mainframe computer.

If one feature or aspect of a mainframe computer was selected to be its defining characteristic then it would undoubtedly be reliable uptime. The vast majority of mainframe computers have provided continual service measured in years and in many instances decades of non-stop functionality.

Redundant Engineering

The major engineering feature of the modern mainframe computer that delivers this degree of reliable service is their considerable amount of redundant internal engineering. This is what gives mainframe computers their high reliability, tight security, extensive input/output facilities, strict backwards compatibility for older software, and high utilization rates (very little processing idle time) to support their characteristic massive throughput capabilities.

Hardware Servicing and Upgrades

In order for a mainframe computer to operate non-stop (run) for many years without interruption all repairs and hardware upgrades can and do take place during the normal operation of the mainframe computer. Once again this is another benefit that the inclusion of internal redundant hardware engineering makes possible.

Performance

Supercomputers; such as those at NASA's Columbia Advanced Computing Facility have their performance measured in terms of the number of floating point operations per second (flops) of which it is capable.

The standard yardstick by which the computational performance of a mainframe computer is measured and subsequently compared with itself at other times or against other mainframes is the number of sustained Millions of Instructions Per Second (MIPS) that it is capable of. As with supercomputers and flops performance the SI prefix system (Mega, Giga, and Peta etc) is also used when stating a mainframes MIPS performance to make these numbers more “human friendly”.

The smallest System z9 IBM mainframes today run at about 26 MIPS while the largest IBM System z10 mainframes can perform approximately 30,657 MIPS (or 30.6 Kilomips).

To give some idea of real world experience, a single mainframe may execute the equivalent of 10 to 100 or even more distributed processors' worth of business activity, however this is highly dependent on the workload. Merely counting processors to compare server platforms is extremely inaccurate.

Multiple Concurrent Operating Systems

Another aspect of the mainframe computer platform that I will only briefly touch on here is their ability to run or host not just one operating system at a time, but many. In this way a single mainframe computer can replace tens or even hundreds of smaller servers. In so doing administrative and management costs are greatly reduced yet at the same time still providing for superior scalability and reliability.

Processing Tasks

Mainframe computer processing has always tended to focus on problems which are limited by input/output and reliability ("throughput computing") as well as solving multiple business problems concurrently (mixed workload). In marked contrast to the supercomputer; which uses massive parallel processing to work on a single highly complex task, the mainframe computer generally makes use of its parallel processing capacity to simultaneously run multiple different less complex concurrent tasks.

Times haven't changed much as the types of tasks that mainframe computers usually perform today still revolve around the so called “mission critical” operations that require much repetitive or parallel processing such as correlation of data collected during a census or a survey, statistical processing and analysis, financial transaction processing (banks) and Enterprise Resource Planning (ERP).

Mainframe Computer Design and Performance Optimization

One of the major critical factors in mainframe processing performance is due to the very nature of the types of tasks that it performs because these tend to involve considerable use of external data sources (input).

Thus; in order to optimize performance, mainframes are built with designs that incorporate numerous ancillary “service” processors whose job it is to supply the main processing core processors with a regulated, steady and persistent stream of data to process and then to service the subsequent output requirements of the main processing core processors processing.

Some of these service processor tasks include cryptographic support, I/O handling, monitoring, notifications, logging, authentication and memory handling. The result is that the total processor count of a mainframe is much higher than would otherwise be obvious from many purely MIPS-based benchmarking measurements as the MIPS-based measurement generally does not include those instructions executed by the ancillary “service” processors just the overall machines productive throughput/output.

One side-effect of this is that adding processors to a mainframe computer will speed up the entire machine's performance over its entire workload transparently.

Fuzzy Marketing

In recent times 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 somewhat misleading as it is widely recognized by the larger players in the mainframe computer industry and academia alike that mainframe computers constitute a class of computer genuinely demonstrably different from all other classes of computational platforms.

Mainframe Pricing

Historically mainframes have earned a reputation for being rather expensive but this is no longer the case. It is now possible to buy and configure a complete IBM mainframe system (with software, storage, and support), under standard commercial use terms, for about $50,000 (U.S.). The price of z/OS starts at about $1,500 (U.S.) per year, including 24x7 telephone and Web support while z10 BC systems start at around $100,000 US.

In addition; many vendors including HP Unisys, HP, Groupe Bull, Fujitsu, Hitachi, and NEC now primarily use commodity Intel CPUs rather than custom processors. This has dramatically reduced their development costs and many have also cut back on their commitment to mainframe software developed for similar reasons some time back and the current economic climate will only add further pressure to continue this trend for sometime into the near future.

Combined these factors all point to a competitive if somewhat stagnant development climate for the mainframe computer with the only real impetus being supplied by IBM who has its own large research and development organization designing their own new, homegrown CPUs; including mainframe processors. IBM is currently expanding its software business; including its mainframe software portfolio.

The company takes the view that with a dramatic reduction in effective competition from its rivals it's open season for them with regards to mainframes. From a future perspective I guess IBM believes that things will improve and when they do they will be in such a dominant position they will not have anything to fear from any potential competition no matter what form it takes.

Conclusion

Above all else it is their reliability that defines and identifies the mainframe class of computer platform with uninterrupted service histories measured in many numbers of years. Many of the current IBM mainframe computers have been working non-stop for over a decade now. Not bad value for the dollar.

The hardware on a network may include: personal computers, mainframes, supercomputers, printers, fax machines, navigational control systems, and interactive entertainment centers.
The software on a network always includes application software, workstation operating systems, and network operating systems.

Networked computers have a number of advantages. They allow information to be exchanged at high speeds, they allow important devices to be shared, and they allow people to connect to their computers over long distances.

Benefit of Networking

Geographically remote areas can be connected to share information. Without actually transferring the entire file to all people involved, several people can simultaneously share large files. Also within a networked environment the information generated by a single user can be shared worldwide instantaneously. This enables faster, more precise communication which should translate into greater accuracy, productivity and cost savings.

• Networking allows different types of computers to communicate. Mac and PC users can share information and resources over a network.
  
• Users on a network can also share physical resources such as scanner, printer, or other expensive piece of hardware. Sharing hardware significantly reduces the expense of running a system.
  

Local Area Network (LAN)

1. Limited to a small geographical region
   
2. Specifically designed to share hardware and software at high speeds.
   
3. Originally developed to connect mainframes to dumb terminals (keyboard and monitor only-no system unit) over 50 years ago.
   
4. Mainframe LANs are faster, more powerful and have higher storage capabilities, while PC based LANs are more flexible to changing environments
   
5. Many companies used a combined network of mainframes and PCs
   
6. Computers
   

Supercomputer Evolution

Supercomputer technologies are evolving just as rapidly as other computer technologies. In fact it is some of these “other” computing technologies that are helping to drive the supercomputer.

During the 1970s all the way through the mid-1980s we saw supercomputers built mainly using vector processors working in parallel. Typically this was anywhere between four to sixteen CPUs.

The next phase of the supercomputer evolution saw the introduction of massive parallel processing and a drift away from vector processors.

Now we find that instead of using “specialist” processors in their design, the supercomputers of today and tomorrow are based on "off the shelf" server-class microprocessors, such as the IBM PowerPC, Intel Itanium, or AMD x86-64.

The modern supercomputer is firmly based around massively parallel processing by clustering very large numbers of commodity processors combined with custom interconnects.

Vector Processing

Vector processing is when the processor takes one instruction and applies it to multiple data or data sets. Vector processing works best when very large data sets are involved. Some vector processing instructions are very complex which saves considerably in instruction decoding time for large data sets but is not necessarily great when it comes to simpler processing that does not involve large data sets.

Because of this modern CPUs have vector processing capabilities built into them where the vector unit runs alongside the main scalar processor and is supplied data by programs that “know” it is there.

Single Instruction, Multiple Data (SIMD)

The modern Graphics Processing Unit (GPU) uses a type of vector processing named Single Instruction Multiple Data (SIMD). This technique saves a lot of processing and processing cycles. Intel's SSE is an example of SIMD processing.

Multiple Instruction, Multiple Data (MIMD)

The processor performs mulitple instructions for vector processing on multiple (vectorised) data sets.

As a matter of interest your average home PC processes more data while you watch a short video than all of the 1970s supercomputers put together.

Current Supercomputer Hierarchal Architecture

The supercomputer of today is built on a hierachal design where a number of clustered computers are joined by ultra high speed network (switching fabric) optical interconnections.

Each cluster member is a computer composed of a number of Multiple Instruction, Multiple Data (MIMD) multiprocessors and runs its own instance of an operating system.

Each of these multiprocessors has multiple processing cores of which the application software is oblivious. These multicore processors share tasks using Symmetric MultiProcessing (SMP) and Non-Uniform Memory Access (NUMA).

Each core is a Single Instruction, Multiple Data (SIMD) processor capable of running a number of instructions simultaneously and many SIMD instructions per nanosecond.

Supercomputer Performance - FLOPS

The performance of “normal” computers is measured in terms of Millions of Instructions Per Second (MIPS). Supercomputer performance on the other hand is measured in terms of Floating Point Operations Per Second.

A floating point number is a number expressed in scientific notation (a basic number, a base and an exponent). For example 4.5546 x 1014. In this example I used the standard scientific notation which uses a base of 10. Binary or base 2 is also used.

With the enormous processing power of a modern supercomputer the number of floating point operations that it executes every second is very high, so we use the SI prefix system to make these numbers more manageable for the human mind.

Mega = 106, Giga = 109, Tera = 1012, Peta = 1015, Exa = 1018 and Zetta = 1021.

The reason why we use floating point notation is that it enables us/the computer to deal with incredably large, long numbers that it would otherwise be unable to do.

The Fastest Supercomputer

The November 2007 edition of the Top500 list placed IBMs Blue Gene/L as the fastest supercomputer running. The Blue Gene/L consists of a cluster of 65,536 computers, each with two processors, each of which processes two data streams concurrently. The IBM Blue Gene/L has a peak processing capacity of 596 teraflops. The Cray XT4 with 101.7 teraflops was second.

IBM Claims one petaflops Blue Gene/P

The chip inside IBM's Blue Gene/P supercomputer consists of four PowerPC 450 cores running at 850MHz each whereas that in the IBM Blue Gene/L had two PowerPC cores running at 700MHz.

Each 2' x 2' Blue Gene/P PCB holds 32 of these quad core PowerPC chips and can crunch its way through 435 billion operations per second. Each 6' rack can hold 32 of these PCBs. The one petaflops IBM Blue Gene/P supercomputer comes with 294,912 processors and takes up 72 racks in all.

IBM says that a three petaflops supercomputer would require 216 of these racks holding 884,736 processors. All of these systems use high-speed optic fiber interconnects.

NASA Enters the Picture

NASA announced that they are planning to build a supercomputer named the Pleiades Project, which they expect will pass the petaflop barrier in 2009 and hit 10 petaflops by 2012. With initial installation of this Intel, quad-core Xeon based machine targeted for completion by July 2008 and initially producing 245 teraflops from a total of 20,480-cores.

This new supercomputer will be installed at NASA's Advanced Supercomputing facility at the Ames Research Center at the Moffett Federal Airfield in California. Moffett is the location of NASA's current supercomputer “Columbia” which became oerational in 2004.

The Columbia supercomputer on the other hand is a cluster of 20 machines, each with 512 processors, each of which processes two data streams concurrently giving it a performance rating of 88.88 teraflops.

The Pleiades supercomputer will be made up of 40 racks, each equipped with 512 processor cores and 512GB of memory. In all the new supercomputer will have more than 20,800 GB of memory. An SGI InfiniteStorage InfiniBand disk solution, designed to store and manage 450 TB of data is also included in the project, which is five times bigger than the entire print collection of the Library of Congress.

NASA stated that they are planning to use the new Pleiades supercomputer for designing a new rocket, modeling, simulation of future missions, hypersonic aircraft, simulate landing deployments and model fabrics for future spacesuits and more.

Relative Processing Performance

A current model Intel quad-core Xeon 2.66 GHz based workstation costing a few thousand dollars now outperforms a 1990s model Cray C90 supercomputer costing many millions of dollars. Many of the cutting edge technologies of the 1990s supercomputer can now be found in your average desktop today.

A supercomputer running at 1 petaflop would outperform a 2.4 Kilometer high stack of laptops.

As of 2007, the fastest PC processors (quad-core) perform over 30 gigaflops but in terms of purely FLOPS performance the Graphics Processing Unit (GPU) in consumer video cards such as those based on the nVidia GeForce 8 series. The 8800 Ultra for example scores 576 gigaflops on 128 processing elements or 4.5 gigaflops per element.

This equates to around 4.5 gigaflops per element. Compare that with the IBM Blue Gene/L's 2.75 gigaflops per core. However, GPUs are to date nowhere as flexible as a general purpose CPU.

Check out the top ten supercomputers on the Top500.org list

Supercomputer Uses Today

Tasks that supercomputers are commonly used for today inlcude calculation intensive tasks such as:

Physics - Quantum mechanics, thermodynamics, cosmology, astrophysics

Meteorology - Weather forecasting, climate research, global warming research,

Molecular Modeling - Computing the structures and properties of chemical compounds, biological macromolecules, polymers, and crystals

Physical Simulations - aerodymanics and fluid dynamics, wind tunnels,

Engineering Design - Structural simulations, bridges, dams, buildings, earthquake tolerance

Nuclear Research - Nuclear fusion research, simulation of the detonation of nuclear weapons

Cryptography and Cryptanalysis - Code and cypher breaking, encryption

Earth Sciences - Geology, volcanology, geophysics

Training Simulators - Advanced astronaut training and simulation

The main users of these supercomputers inlcude: universities, military agencies, scientific research laboratories and major corporations.