Super Computers are the fasted computers we know of. They are characterized by very high computational speeds and a immense number of processors. They are usually owned by countries or corporation and never for personal use. In fact a supercomputer can simply fill a big room.
A supercomputer is a computer at the frontline of current processing capacity, particularly speed of calculation. [Wikipedia]
How do they build them?
Our normal computers have a single processor with multiple cores. Supercomputers are designed with a large number of processors.
Lets consider a team that wants to build a supercomputer.
- First they have to choose the types of processors they want to use. The processor manufacturers are no strangers, Intel, AMD or nVidia.
- The team also chooses the amount of RAM they have to use.
- Now the processors and RAM are inserted into nodes. A node a may contain different types of processors. The type and number of processors, the amount of RAM is design dependent.
- These nodes are interconnected forming a Blade.
- Now the blades themselves are interconnected and stacked up in Racks (Cabinets). Each Rack maybe a size of a large refrigerator.
- Now, a number of these Racks are interconnected in a room, forming the processing component of a supercomputer.
- A display might be attached, but a lot of other computers or dummy terminals are also attached to the supercomputer.
- Every supercomputer runs a specially tailored OS for its specific needs. Now, Linux is used in most supercomputers.
How fast are they?
Just like we measure the speed of a car in kilometers per hour (kmph), we measure speed of supercomputers in FLOPS or Floating-point Operations Per Second. Simply, floating point operations means computations that involve very large decimal numbers, usually 300 digits in a single number.
So FLOPS measures the number of ‘floating point instructions’ a supercomputer can perform or complete in one second. But FLOPS is a basic unit, just like bytes is a basic unit of memory. We use gigabyte(GB), terabyte(TB) or even petabyte(PB) to indicate memory size. We use petaFLOPS to measure the speed of the present supercomputers.
A petaFLOPS means one quadrillion operations per second.
A quadrillion is 1 followed by 15 zeros.
Lets see the 10 fastest Supercomputers in the world. (Nov 2012)
|8||Tianhe – 1A||2.57||China|
|10||DARPA Trial Subset||1.51||USA|
Why super computers?
Supercomputers are mainly used for tasks which require immense computational power. Such tasks, if attempted on the normal computers will take an incredibly long time, sometimes several years to complete.
Following tasks require a supercomputer :
- problems including quantum physics
- weather forecasting
- climate research
- oil and gas exploration
- molecular modeling
- physical simulations
The above tasks are so computationally intensive its practically impossible to complete them using normal computers.
Supercomputers in action
Supercomputers are being used in the most unique ways. They are mostly concentrating on the key issues on technology, medicine, defense equipments etc. Lets see some of the supercomputers in action.
The fastest supercomputer in the world is used to study alternative energy resources and climate change simulations in a global level.
This one’s a life saver. USA doesn’t have to actually conduct nuclear tests anymore, Sequoia simulates them. This one makes sure that the nuclear weapons are ready for action any time of the day.
It works on atom and electron simulations useful for solar cells, effects of tsunami on buildings, building better structures in earthquake prone zones. It is also used for pharmaceutical research and semi-conductor technology.
Mira by IBM, simulates the evolution of the Universe. It also runs the climate change scenarios and helps in research of more efficient car batteries.
Tianhe-1A, appropriately translated means ‘Milky Way – Number One’. When it was introduced, it is the fastest supercomputer in the world. Aptly for its name, it is used for studying the formation of galaxies in the Universe. Also its helps in the research in hurricane and tsunami modeling, cancer research and drug discovery.
Watson, IBM’s very own supercomputer is used in Cancer in a collaboration with Memorial Sloan-Kettering Cancer Center. Watson can process and understand natural language, language that we speak. So it can easily manage all the research data available and suggest relevant material to a particular researcher.
What Watson can do is read and understand huge volumes of information. There is so much information being developed in health care in general, and oncology in particular, that the ability to understand all the information out there is becoming progressively more challenging. What Watson does is bring information to the doctor. [Dr. Martin Khon, Chief Medical Scientist, IBM]
Supercomputers in India
- India has 9 supercomputers.
- Notable one’s – SAGA-220, EKA, Virgo, PARAM Yuva, Prithivi and the latest PARAM Yuva-II.
- In 2007, EKA was the worlds 4th fastest according to TOP500.
- The fastest supercomputer from India that is listed in TOP500 is located at ‘Center for Mathematical Modeling and Computer Simulation’. It is not named.
- Later India developed PARAM Yuva II on February 8, 2013. Now it is the fastest supercomputer in India.
- By 2017, India plans to build the fastest supercomputer in the world.
- TOP500 is a project that lists and details the fastest 500 supercomputers in the world.
- World’s first supercomputer is CDC 6600 by Cray.
- The fastest supercomputer is Titan. (TOP500)
- At least 50% of TOP500 are from USA.
- Almost 74% use Intel Xenon architecture for processors.
- Most of them use water coolant systems.
The number of supercomputers in the world increases rapidly. They are sometimes seen as an indication of a country’s technological advancement and surely they are. There is an estimation that in future, personal computers can run 10 times more faster than today’s supercomputers. But a new technology, Exascale computing which would make supercomputers run a 1000 times faster than their present counterparts.
Our quest of faster and more efficient computers is insatiable. We will keep on building systems that are better than the last one. Maybe some day we can build the kind of supercomputers that will simulate the human brain. That would be the holy grail of Artificial Intelligence. And our scientists say that it is not so far.