User:Alan Cleary/CIS 300/presentation
What is a Supercomputer
The term "super computer" is very liquid and can have many literal meanings. Roughly, a supercomputer is a computer that is at the frontline of current processing capacity, particularly speed of calculation. They consist of hundreds, sometimes thousands of processors designed for specific types of calculations/applications involving quantum physics, weather forecasting, climate research, molecular modeling, physical simulations, and nuclear fission.
Super computers were originally introduced in the 1960's and were primarily designed by Seymour Cray at CDC. This lead to the supercomputer market in the 1970's. Cray later left the CDC and created his own company named Cray Research that dominated the super computer market from the mid 80's into the 90's. After the introduction of the minicomputer in the 80's many smaller competitors entered the market but were lost in the "super computer market crash" in the 90's. Today the leading super computer companies are Cray, IBM, and Hewlett Packard. In May of 2010 the Cray Jaguar, with 224,256 x86-based AMD Opteron processor cores, was declared the fastest super computer in the world.
The most common super computer today is what's know as a cluster computer. This means the computer is actually many computers running linearly as one through some kind of framework, such as a Local Area Network.
The processors are generally custom made focusing on a specific type of computation such as numeric computations. One of the primary concerns while designing a super computer is the memory hierarchy. Unlike traditional desktop computers, the designers want the processors to have a continuous feed of data and instructions. This allows for a much higher bandwidth with latency being less of an issue because super computers aren't generally used for transactional processing.
The operating systems used today are generally variants of Linux.
The software used for distributed processing includes APIs such as MPI, PVM, VTL, and open-source solutions such as Beowolf, WareWulf, and OpenMosix, which enables the creation of a supercomputer from workstations and/or servers.
Cray Jaguar (XT5)
AMD x86_64 Opteron Six Core 2600 MHz (10.4 GFlops) Total combined memory amounts to over 360 terabytes
The Jaquar is often used for problems in areas such as climate modeling, renewable energy, materials science, seismology, chemistry, astrophysics, fusion, and combustion.
At the 2010 National Conference on High-Performance Computers (HPC)in China, China unveiled the Tianhe-1A, meaning Milky Way (literally Sky River) one A in English.
The $88 million computer is currently stored at the National Supercomputing Center of Tianjin. It's theoretically able to do more than 1 quadrillion calculations per second (one petaflop) at peak speed. Tianhe-1A's peak performance reaches 1.206 petaflops, and it runs at 563.1 teraflops (1,000 teraflops is equal to one petaflop) on the Linpack benchmark.
- http://en.wikipedia.org/wiki/Supercomputer and sub-references