The Race for PFLOPS is the New Race for Space

2008's 2nd place supercomputer, Cray's Jaguar.  Image courtesy of Wired.

Supercomputers are ones that are used for extremely complex processing jobs such as problems in quantum physics, weather forecasting, climate research, molecular modeling, and physical simulations, to name some of the better uses.  They were introduced in the 1960s, and were initially designed by Seymour Cray.

1985 Cray-2 seen from the side at the Musée des Arts et Métiers in Paris.
Image courtesy of Edal Anton Lefteroy/Wikipedia.

LINPACK is a software library for performing numerical linear algebra on digital computers, designed for use on supercomputers in the 1970s and early 1980s.  The LINPACK Benchmarks are a measure of a computer's floating point power, measuring how fast a computer can solve a dense system of linear equations.  The result is reported in millions of floating point operations per second (FLOPS). Supercomputers are currently measured in PFLOPS, or petaFLOPS.

IBM Roadrunner, the world's first PFLOPS computer.
Image courtesy of LeRoy Sanchez/Wikipedia.

The TOP500 Project ranks and details the 500 most powerful computer systems in the world (non-distributed computers, that is, or one that does not interact with other computers to solve a complex problem).  It began in 1993 and publishes an updated list twice a year, in June and November.  The June list coincides with the International Supercomputer Conference, and the November one with the ACM/IEEE Supercomputing Conference.  The Project's goal is to provide a realiable basis for finding and tracking trends in high-performance computing.  It bases its rankings on HPL, a software package regarded as a portable implementation of of the LINPACK Benchmark, which is written in Fortran.

The list is compiled by Hans Meuer of the University of Mannheim in Germany, Jack Dongarra of the University of Tennessee, and Erich Strohmaier and Horst Simon of NERSC/Lawrence Berkeley National Library.

TOP500 June 2011 Poster

This month the TOP500 Project committee announced that Fujitsu's K supercomputer was the fastest supercomputer in the world with a performance of 8.162 PFLOPS, or 8.162 quadrillion calculations per second with 93% efficiency. (A quadrillion is 1,000,000,000,000,000 or one thousand million million or  10¹⁵.)

The K supercomputer.  Image courtesy of The Telegraph, UK.

Named after the Japanese word "kei", which stands for 10 quadrillion, the supercomputer is produced by Fujitsu and located at RIKEN Advanced Institute for Computational Science campus in Kobe, Japan.  According to Jack Dongarra, the K's performance equals "one million linked desktop computers".  Its power usage is roughly that of 10,000 houses and its annual running costs are $10 million (U.S.).

A rack of RIKEN's next-generation supercomputer manufactured
by Fijitsu.  Image courtesy of CES1596/Wikipedia.

The previous winner, announced just last November, was China's Tianhe-1, which was capable of 2.566 PFLOPS.  Located at the National Supercomputing Center of Tianjin, China, it was developed by the Chinese National University of Defense Technology (NUDT) in Changsha, Hunan.

The Tianhe-1A CPU-GPU hybrid supercomputer.  Image courtesy of The Register, UK.

The rankings reflect how quickly computer power is advancing, hence the new ratings every six months.  Asian countries have made huge investments in supercomputing and lately dominate the field.  There is, however, a contender called Blue Waters being developed at the University of Illinois at Urbana-Champaign which may soon rival the K in speed, per Dongarra.

Radio Shack's Tandy TRS89 computer system.
Image courtesy of Wikipedia.

As someone commented, if you bought a Tandy TRS80 in 1981, the K computer is 2,050 trillion times faster.  This is great advancement but it all comes down to what it will be used for.  If it used for something trifling like stock predictions, or something heinous like weapons development it would be a shame.  But if it used in forecasting or in the field of medicine, things that will benefit the health and safety of mankind, then this is truly an amazing development.

***************

Check out RIKEN to find out more about the K.

Geeks can learn more about programming on the K.

*******************************

From Paper Cranes to Mad Cow Disease

Green bottle by Martin Demaine.

Martin Demaine is an artist and mathematician who is currently an artist in residence at MIT (Massachusetts Institute of Technology).  After graduating from high school in Massachusetts, he went to England to study glassblowing. Eventually he opened up shop in New Brunswick where his was the first one-man glass studio in Canada.  An excellent craftsman, his work ended up in various major museums.

Martin Demaine at work.

In 1981, he was involved in a very different creation:  his son Erik was born. When Erik was six years old, Martin and Erik formed a puzzle company and distributed their puzzles throughout Canada.  Erik was home-schooled and earned his B.S. at the age of 14. By the time Erik was 20, he had his PhD and was a professor at MIT - said to be the youngest professor MIT ever hired.  In 2003, Erik was awarded a MacArthur Fellowship, the so-called genius award, as a computational geometer.

Erik Demaine.  Photo by William Plowman.

His PhD dissertation was seminal in the field of computational origami, and was award-winning in itself.  Together, father and son are jointly engaged in works that involve both mathematics and art, still focusing on the computational complexity of games and puzzles, among other things.

Computational Origami by Martin and Erik Dumaine,
in the permanent collection of MoMA.

In a post I did earlier on origami, I briefly touched on computational origami and mentioned Robert Lang, an earlier pioneer in the field.  This is a type of computer program for modeling the ways that different materials, especially paper, can be manipulated.  Besides amazing origami pieces, such as insects complete with antennae, there is a more practical use (and potentially a lifesaving one).

Origami insects by Japanese artist Taketori, courtesy of his website.

Computational origami basically applies to engineering problems where large surfaces need to be fitted into small or flat spaces without cutting them, just by folding.  For some of us it would facilitate the refolding of a road map.  More importantly, it would facilitate folding airbags - the ideal folds would allow the airbag to function correctly yet take up little space.  Even more importantly, this concept can be applied to computer processors, fitting an enormous amount of data onto the smallest possible area.

But perhaps the most important of all potentialities is studying folds in protein, which Erik Demaine is working on.  Computational origami could ascertain if proteins have "bad folds" and could help crack the secrets of protein structure and sequences.  This could lead to cures for Alzheimer's, cystic fibrosis, emphysema, many cancers, and even mad cow disease.

Basic protein folds, image courtesy of www.cellbiol.net.

However, to compute these complexities requires a computer operating at something like a quadrillion operations per second (1 petaflop, or 1,000 teraflops). IBM has been working on some supercomputers in its Blue Gene project.  Blue Gene is a computer architecture project that explores the production of supercomputers designed to be able to operate in speeds of the petaflops range. Thus far they have reached a peak speed of 596 Teraflops.

A Blue Gene/P supercomputer at Argonne National Laboratory.
Image courtesy of Wikipedia.

Once the right supercomputer is built, it may still take decades to learn protein folds and their applications to health issues.  But the possibilities are amazing. Today Erik Demaine is a part of the MIT Computational and Systems Biology Initiative (CSBi) which "links biologists, computer scientists and engineers in a multi-disciplinary approach to the systematic analysis of complex biological phenomena".  Martin Demaine is also a visiting scientist and works on computational glassblowing.  From art comes science...

***************

Unless otherwise stated, images from the websites of Martin and Erik Demaine.  

Please check these excellent websites out for more information and clarification:

Martin Demaine

Erik Demaine

*******************************