| |
History of computers 1959 to 2004
|
|
|
| |
| About 50 years ago, the Synoptic and
Dynamical Research Sub-committee of the Meteorological Research
Committee recommended that the Met Office obtains an electric desk-calculator,
and recruit a mathematician specially trained in computational methods,
to enable the Office to undertake trial meteorological computations.
So it was, with such basic equipment, not even capable of automatic
multiplication, that the Met Office began to fulfil the dream
of L.F. Richardson, when in 1922 he suggested the possibility
of numerical weather prediction using a grand 'forecast factory'
- consisting of thousands of (human) computers, seated around
a globe-shaped auditorium, each one calculating the future state
of the atmosphere at a fixed location by evaluating a set of equations.
In the ten years following the purchase of the electrical desk-calculator,
a small number of Met Office staff had access to a 'real' computer
- the LEO 1 - which had been built by Lyons, the caterers. Eventually,
in 1959 the Met Office purchased its first computer - a Ferranti
Mercury, which the office named Meteor. With glowing valves and
copious volumes of paper tape, the Met Office had entered the
computer age.
|
| |
|
|
| |
|
Through the decades following the arrival of Meteor, the Met
Office has regularly upgraded and replaced its computers, to take
advantage of the speed and functionality provided by new technology.
During this time, the size and complexity of the meteorological
models and their associated data have also increased dramatically,
reflecting the steady increase in computational power and memory
sizes. The table below shows the increase in computational power
of the Met Office's computers over the last 35 years, and how
the meteorological models have also increased in both horizontal
and vertical resolution during this period.
|
Met Office computers from 1959 to
2004 |
| |
Computer |
Calculations
per second |
Main memory (words) |
Horizontal
resolution
(Global/local)/levels |
| 1959 |
Ferranti Mercury |
3 x 103 |
1 x 103 |
(N.A./320 km)/2 levels |
| 1965 |
English Electric KDF 9 |
50 x 103 |
12 x 103 |
(N.A./300 km)/3 levels |
| 1972 |
IBM 360/195 |
4 x 106 |
250 x 103 |
(300 km/100 km)/10 levels |
| 1982 |
CDC Cyber 205 |
200 x 106 |
1 x 106 |
(150 km/75 km)/15 levels |
| 1991 |
Cray Y-MP C90/16 |
1 x 109 |
256 x 106 |
(90 km/17 km)/19 levels |
| 1997 |
Cray T3E 900/1200 |
1.5 x 1012 |
36 x 109 |
(60 km/12km)/38 levels |
|
|
| |
|
|
| |
|
Between 1997 and 2004, the Met Office used two Cray T3E supercomputers.
The T3E was a radical departure to what had been used previously.
Instead of having a single processor like the IBM 360/195 or just
several like the Cray Y-MP, it was capable of having several hundred
processors. This type of computer is known as MPP or Massively
Parallel Processor. The T3E used commodity ('off-the-shelf') processors
rather than specifically designed ones like the C90 did and thus
despite containing many more processors than the C90, it was not
proportionally more expensive.
|
| |
|
|
| |
|
In 2004, following successful relocation to its new headquarters
building in Exeter, the Met Office changed computers again. Initially
the operational forecast was run on a NEC SX-6 supercomputer but
was transferred to a more powerful NEC SX-8 supercomputer in April
2005. The SX-6 still remains key to the Met Office's capabilities;
in addition to acting as a back-up machine, the SX-6 will be used
for research and development.
The SX-6 and SX-8 differ to the T3E in that they have fewer,
but much more powerful processors. The SX-6 and SX-8 machines
are divided into 'nodes' with each node containing eight processors.
Currently the SX-8 has 16 nodes, and there are two SX-6 clusters;
one with 19 nodes, and the other with 15 nodes. Each SX-8 node
is twice as powerful as an SX-6 node, so together, the combined
systems deliver over thirteen times the sustained power of the
previous Cray T3E computers, which the SX-6 replaced in 2004.
Weather forecasts will benefit from the extra power, which will
mean increased accuracy and detail in the models run by the supercomputer.
These models will provide improved forecasts of high-impact weather,
building on data available from a new generation of satellites.
Climate change predictions will become even more authoritative
through increases in resolution; representation of new processes
and the use of ensemble predictions to provide risk assessments.
|
|
|
|