Hardware and Software

Mainframe computer
A mainframe is a multitasking, multi-user computer, meaning it is designed so that many
different people can work on many different problems, all at the same time.

A mainframe computer can have hundreds of processor cores and can process a large
number of small tasks at the same time very quickly.

Mainframes are much larger and have more processing power than personal computers,
laptops or tablets and cost considerably more to buy.

Mainframe computers are used mainly by large organisations for bulk data-processing
applications.

Mainframe computers have almost total reliability, being very resistant to viruses and
Trojan horses.

The mainframe is able to process more transactions and calculations in a set period of
time when compared with alternatives and it continues to operate with a minimum of
downtime, which means that companies can operate 24 hours a day, every day.

Super computer
Supercomputers can have in excess of 100 000 processing cores.

CPU Cores
Most PCs and laptops used to have a single processor, but today they tend to have a
CPU with many processor cores which gives the effect of having many processors.

The number of tasks that a PC can do simultaneously depends on the number of

processor cores it has.

For example, if a PC has a processor with 20 cores, it means it can carry out 20 tasks at
the same time.

Serial processing & Parallel processing

Serial processing is when a computer performs tasks one at a time.

Parallel processing is when a computer performs several tasks simultaneously.

1
 Characteristics of a Mainframe
Longevity

Mainframe computers have great longevity, or lifespans. They can run continuously for
very long periods of time.
Many of the current mainframes have existed for decades and are still working well.
To shut mainframes down and then dispose of the hardware is very expensive, as is the
hiring of companies to securely remove their data.
Rather than replace their mainframes, companies using them are more inclined to
modernise them using different software tools.

Threats to longevity
1) Lack of IT professionals
IT professionals who can maintain or program mainframes are in shortage. Many
of the older mainframes still work on the COBOL programming language, which is
being replaced by Java, C and Python. Yet, 95% of ATM machines still use
COBOL.

2) Cloud computing
Cloud computing is accessible from anywhere, thereby reducing the need to
maintain expensive hardware within an organisation.

RAS (Reliability Availability Serviceability)

Reliability
Mainframes are the most reliable computers because their processors are able to check
themselves for errors and are able to recover without any undue effects on the
mainframe’s operation.
The system’s software is also very reliable, as it is thoroughly tested and updates are
made quickly to overcome any errors.

Availability
Mean time between failures (MTBF) is a common measure of systems. It is the average
period of time that exists between failures (or downtimes) of a system during its normal
operation.

Mainframes are available at all times and for extended periods.

Mainframes give months or even years of system availability between system downtimes.
In addition to that, even if the mainframe becomes unavailable due to failure, the length of
time it is unavailable is very short.
It is possible for a mainframe to recover quickly, even if one of its components fails, by
automatically replacing failed components with spares. Spare CPUs are often included in
mainframes so that when errors are found with one, the mainframe is programmed to
switch to the other automatically. The operator is then alerted and the faulty CPU is
replaced, but all the time the system continues to work.

2
Serviceability
This is the ability of a mainframe to discover why a failure occurred and means that
hardware and software components can be replaced without having too great an effect on
the mainframe’s operations.

Security
Mainframes have greater security than other types of computer systems.

A mainframe has many layers of security including:

 user identification and authentication, although more and more systems are using
multi-factor authentication, which is a combination of two or more of the following: a
password, a physical token, a biometric identifier or a time-restricted randomised PIN
 levels of access, which means that it depends on a user’s level of security as to which
sets of data they can access
 encryption of transmitted data and data within the system
 secure operating systems
 continual monitoring by the system for unauthorised access attempts.

Performance metrics
The performance metrics of a computer are the measures used to determine how well,
or how fast, the processor deals with data.

The speed of a mainframe’s CPU is measured in millions of instructions per second

(MIPS).
However, it is not always the best measure, because not all instructions are the same.
Mainframes use a very large number of different instructions, with some being
straightforward and easy to carry out, while others can be more complex and slower to
process.
An application using five million simple instructions will take a lot less time than one using
five million complex ones.

The comparison between the performance of one mainframe and another should be
made by measuring how fast the CPUs are when carrying out the same task.
This is referred to as a benchmark test.
This measure is usually obtained when processing application software.

MIPS are often linked to cost by calculating how much a mainframe costs per one million
instructions per second.

Volume of input, output and throughput

Mainframes have specialised hardware, called peripheral processors, that deal
specifically with all input and output operations, leaving the CPU to concentrate on the
processing of data.

This allows mainframes to deal with very large amounts of data being input (terabytes or
more) and subsequently very large volumes of output being produced.

3
Modern mainframes can carry out many billions of transactions every day. This large
number of simultaneous transactions and extremely large volumes of input and output in
a given period of time is referred to as throughput.

Fault tolerance
A computer with fault tolerance means that it can continue to operate even if one or more
of its components have failed.
It may have to operate at a reduced level, but does not fail completely.

Mainframe computers are fault-tolerant in terms of their hardware.

While in operation, if a processor fails to function, the system is able to switch to another
processor without disrupting the processing of data.

The system is also able to deal with software problems by having two different versions of
the software. If the first version produces errors, the other version is automatically run.

Operating system
Most mainframes run more than one operating system (OS) at any given time.
The use of z/OS, z/VM and Linux (which are all ®, ® different operating
systems) at the same time often occurs.

The OS on a mainframe divides a task into various sub-tasks, assigning each one to a
different processor core.
When each sub-task has been processed, the results are recombined to provide
meaningful output.
This is called parallel processing.

Number of processors
Early mainframes had just one processor (the CPU), but as they evolved
more and more processors were included in the mainframe system.

The number of processor cores found in a mainframe is now measured in the hundreds.

Heat maintenance
More powerful systems produce more heat.
The large amount of heat produced by a system also has an effect on the lifetime of
components other than the processors.

For mainframes, it is recommended that liquid-cooling systems (water-cooled) be

used rather than air-cooling systems (fans).

4
 Characteristics of a Super computer
Longevity
Supercomputers have a lifespan of about five years.
Research institutions and meteorology organisations are always looking for faster ways to
process their data and so, unlike companies using mainframes, will tend to look at
replacing their existing systems whenever much faster supercomputers come on to the
market.

Security
Most supercomputers use end-to-end encryption, which means that only the sender or
recipient is able to decrypt and understand the data.

Performance metrics
The performance metrics of a computer are the measures used to determine how well,
or how fast, the processor deals with data.

Supercomputers use a different set of metrics. As they are used mainly with scientific
calculations, their performance is measured by how many Floating point Operations can
be carried out Per Second (FLOPS).

One flop per second means that the calculating device is performing one arithmetic
operation per second.

Speeds of supercomputers are measured in petaflops.

One petaflop is 1 000 000 000 000 floating point operations per second.

Volume of input, output and throughput

A supercomputer is designed for maximum processing power and speed, whereas
throughput is a distinct mainframe characteristic.

Fault tolerance
Supercomputers have far more components than a mainframe, with up to a thousand
times more processors alone.

The approaches to fault tolerance are much the same as those for mainframe
computers, but with millions of components, the system can go down at any time, even
though it tends to be up and running again quite quickly.

Operating system
Supercomputers tend to have just one OS (Linux).

But most supercomputers utilise massively parallel processing in that they have many
processor cores, each one with its own OS.

Number of processors
Supercomputers have hundreds of thousands of processor cores.

5
Unlike mainframes, modern supercomputers use more than one GPU or
graphics processing unit.

Heat maintenance
More powerful systems produce more heat.
The large amount of heat produced by a system also has an effect on the lifetime of
components other than the processors.

Supercomputers always have had an overheating problem.

Some supercomputers draw four or more megawatts of power to keep them operating at
high efficiency. That is enough to power several thousand homes.
This, together with having so many processors very close together, results in a great deal
of heat being produced and requires the use of direct liquid-cooling to remove any excess
heat.

6
 Uses of Mainframes
Mainframes are usually used in
 Banks
 Finance companies
 Insurance companies
 Energy providers
 Health care providers
 Airlines
 Travel agencies

Mainframes are also used in areas such as

 Censuses
 Industry statistics
 Consumer statistics
 Transaction processing

Census
Census is an official survey of the people and households in a country that is carried out
in order to find out how many people live there.
The amount of data that has to be processed is enormous.
The processing of census data has long been associated with mainframes.

Industry statistics
Industry statistics are statistics that are recorded regarding trends in different industries,
such as those that process raw materials, make goods in factories, provide services, etc.

Industry statistics can include the number and names of businesses, the number of
employees, wages paid, etc.

Some businesses in certain sectors of industry need mainframes to process this vast
amount of data.

Consumer statistics
Consumer statistics allow businesses to
 assess the demand for their product, that is how many people need or want that
type of product
 know the range of household incomes and employment status of those consumers
who might be interested in the product so that a price can be set
 know where the consumers live for local sales or how inclined they are to use the
internet for shopping
 know how many similar products are already available to consumers and what
price they pay for them

These statistics produce an incredible amount of data and the organisations that produce
these statistics are likely to use mainframes.

7
Transaction processing
Transaction processing can consist of more than one computer-processing operation,
but these operations must combine to make a single transaction.
Each of the operations that constitute the transaction must be carried out without errors
occurring.
In the event of an error, a transaction processing system will remove all traces of the
operations and the system will continue as though the transaction never happened.
However, if there are no errors, the transaction is completed and the relevant database is
updated before the system continues on to the next transaction.
All this has to happen in milliseconds, as it must seem to the user that their transaction
has been carried out immediately.

Why is a mainframe suitable for transaction processing?

 The system must be able to perform millions of transactions in a second.
 The system must be continuously available during the times when users are
entering transactions.
 The system must be able to deal with hardware or software problems without
affecting the integrity of the data.
 The system must be able to add, replace, or update hardware and software
components without shutting down the system.

A mainframe is suitable for transaction processing because

 It can perform hundreds of MIPS.
 It can also transfer processing from one core to another.

8
 Uses of Supercomputers
Supercomputers are used in
 National defence
 Drug research
 Genetic analysis

Quantum mechanics
Quantum mechanics is the study of the behaviour of matter and light on the
atomic and subatomic scale.
Quantum mechanics require a very large number of calculations which
require great accuracy and thus require the use of a supercomputer.

Weather forecasting
Weather forecasting is based on the use of very complex computer models.
Using the past weather readings, the computer examines similar patterns of weather to
those being experienced at the moment and is able to predict the resulting weather.
Data from the sensors at weather stations around the world is input to the
model and then many calculations are performed.
These readings, together with observations from radar, satellites, soundings from space,
and information from ships and aircraft, help the supercomputer to produce a 3-D model
of the Earth’s atmosphere.
Because of the complexity of the calculations, and the very large number of them that
need to be carried out, they can only run effectively on supercomputers.

Climate research
Climate is measured over a long timescale.
The data collected over several decades can be used to show the trends of different
variables over time.
The planet Earth can be regarded as a complex system which has many different
constituent parts. The interactions of each of these
components within the Earth system can be described in mathematical terms.
However, there are very many variables which have to be collected and complex
calculations which need to be carried out.
We have to rely on computer models to help us understand how the system works.
Climate is one component of this planet’s system and only supercomputers are able to run
the models that represent the interactions between the components of this system.

9
 Advantages of Mainframes
1) A mainframe rarely has any system downtime.
2) A mainframe can store and process huge amounts of data.
3) A mainframe can run different operating systems.
4) A mainframe can cope with data coming in a variety of database formats.
5) A mainframe has stronger security than other systems. It uses complex encryption
systems and authorisation procedures.
6) A mainframe is scalable because processors and memory can be added as required.
7) Hardware and software upgrades can occur while a mainframe is still up and running.
8) Mainframes are getting faster and more powerful every year.

Disadvantages of Mainframes
A mainframe is very expensive to
1) Buy and can only be afforded by large organisations such as multinational banks.
2) Run and maintain as there is a high cost for the personnel required to do that.
3) Run as the software required to run mainframe systems is costly.
4) House as large rooms are required.
5) Cool as the cooling systems needed become costlier to install and run.

Advantages of Supercomputers
1) Supercomputers are the fastest data processing computers.
2) Supercomputers use massively parallel processing, which makes them more powerful
compared to the parallel processing of mainframes.

Disadvantages of Supercomputers
1) Supercomputers are the most expensive computers to buy and install.
2) Most supercomputers have one operating system.
3) Supercomputers are less fault tolerant than mainframes.
4) supercomputers are less likely to recover as quickly in the event of the failure of one
component.
5) Supercomputers are down more often than mainframes.

10