Top500 Supercomputers

BlueGene/L system, No.1 Supercomputer [Photo courtesy: Lawrence Livermore National Laboratory (LLNL)]

The 29th edition of the TOP500 list of the world's fastest supercomputers was released at the International Supercomputing Conference held in Dresden, Germany. For the fourth straight time, the top spot was occupied by the BlueGene/L system, developed jointly by IBM and Department of Energy (DOE)'s National Nuclear Security Administration (NNSA), and installed at DOE's Lawrence Livermore National Laboratory in Livermore, California. The BlueGene/L reached a Linpack benchmark performance of 280.6 TFlop/s (“teraflops” or trillions of calculations per second).

While the No. 1 system is still unchallenged, the rest of the TOP10 experienced large changes since November 2006. In this year's list, two other systems exceeded the level of 100 TFlop/s: the upgraded Cray XT4/XT3 at DOE’s Oak Ridge National Laboratory, ranked No. 2 with a benchmark performance of 101.7 TFlop/s; and Sandia National Laboratory’s Cray Red Storm system, which ranked 3rd at 101.4 TFlop/s.

Two new BlueGene/L systems entered the TOP10. They are both located in the state of New York and represent the largest academic supercomputer installations. The No. 5 system is installed at Stony Brook, NY at the New York Center for Computational Science (NYCCS). The No. 7 system is at the Rensselaer Polytechnic at the Computational Center for Nanotechnology Innovations (CCNI), Try, NY.

The fastest supercomputer in Europe is an IBM JS21 cluster at the Barcelona Supercomputing Center in Spain, which ranked No. 9 at 62.63 TFlop/s. The highest ranked Japanese system is located at the Tokyo Institute of Technology and ranks No. 14 on the list. This system is a cluster integrated by NEC based on Sun Fire x4600 with Opteron processors, ClearSpeed accelerators and an InfiniBand interconnect.

Mathematics Explains Genetic Instability in Cancer

Natalia Komarova (photo courtesy: UC Irvine)

Natalia Komarova is an associate professor at the University of California, Irvine. One of her research interests lies in developing mathematical models to study initiation and development of cancer, viewed as somatic evolution in populations of cells. She has investigated the following crucial issues related to the understanding of cancer: the role of genetic instability in cancer progression, cellular origins of cancer, the role of stem cells in carcinogenesis, and even questions like 'how can we fight resistance to drug therapies'? Natalia formulated important questions of cancer biology in the language of mathematics, and used experimental data to validate her mathematical models.

In a recent publication in the Journal of the Royal Society Interface, Natalia and her colleagues Alexander Sadovsky and Frederic Wan of UC Irvine provided a good insight into how cancerous tumors thrive and formulated a potential foundation for future cancer treatments. Their research focused on the phenomenon of genetic instability, a common feature of cancer in which cells mutate at an abnormally fast rate. Such mutations can cause cancer cells to grow, or they can cause the cells to die.

The UC Irvine team tried to find an answer to a question: How can a tumor optimize its own growth? They formulated an optimal control problem for the mutation rate in cancer cells and then developed a method to find optimal time-dependent strategies. The results from a wide range of parameters showed that cancerous tumors grow best in an early 'unstable' stage of development when cancerous cells mutate to speed up malignant transformation. The growth stabilizes at later stages by turning off the mutation rate, a fact that agrees very well with the growing biological evidence for such phenomena. They also succeeded in identifying parameter regimes where it is advantageous to keep the state stable (or unstable) constantly throughout the growth.

Previous studies have observed such genetic pattern by using laboratory techniques, but the UC Irvine results could explain for the first time by using a mathematical model why this pattern leads to tumor growth.

Reference:
"Selective pressures for and against genetic instability in cancer: a time-dependent problem"
Natalia L. Komarova, Alexander V. Sadovsky, Frederic Y.M. Wan
The Journal of the Royal Society Interface Link to Abstract

High Performance Computing to Facilitate Analysis of Brain

Scientists at CUBRIC analyzing MRI scan data (photo courtesy: CUBRIC, Cardiff Univ., UK)

The Cardiff University Brain and Repair Imaging Centre (CUBRIC) is one of the first facilities in the UK to combine Structural and Functional Magnetic Resonance Imaging (MRI/fMRI) and Magnetoencephalography (MEG) and is devoted to pioneering advanced brain scanning techniques (fMRI/MEG) capable of mapping the structure and function of the healthy and impaired brain.

Last week, scientists at CUBRIC announced that they could use high-performance computing to analyze brain scans 24 times faster than was previously possible, giving a more complete view of the brain.

The technique overlays two types of magnetic images, magneto encephalography and magnetic resonance imaging scans, to provide single, comprehensive examination of the brain. Using the 300-node cluster, scientists can analyse 100 complete brain images in just 16 minutes. The analysis involves use of complex mathematical techniques on data sets and in the past that would take significant time. Now, it's possible to take a deeper look into a complex organ and how it reacts to adverse conditions.

Scientists have also found that the processing power has allowed them to achieve more in each visit, reducing the overall number of costly visits necessary for their research.

Manchester Centre to Study Complexity of Real Life

Sackville Street Building of the Faculty of Engineering and Physical Sciences (photo courtesy: University of Manchester)

The Centre for Interdisciplinary Computational and Dynamical Analysis (CICADA) of the Faculty of Engineering and Physical Sciences at the University of Manchester is developing a £3m research centre to provide cutting-edge research on complex, real-life systems such as car safety systems, flight controllers, power stations and the human body. The first projects are due to begin in October.

Existing programs break down when analysing these systems, as they cannot cope with the complexity of continuous change. Unfortunately, many safety devices, such as electronic stability control in cars, have to rely on such systems, where failure could result in loss of life. The centre aims to bring biologists, mathematicians and engineers together to help understand problems that combine both discrete and continuous data. It is hoped the research will help to unlock the potential of existing, technologically advanced systems, that haven’t been integrated successfully into real-life situations because we cannot predict their behaviour accurately enough.

The research could also provide insights into biology, where continuous changes in chemical concentrations result in discrete changes in cells. Discrete state changes in the brain and cells are triggered by continuous changes in chemical concentrations. In biological systems there are generally many of these processes interacting in large complex science networks and these are hard to analyse using mathematical or computer science approaches that currently exist.

CICADA aims to attract internationally renowned scientists and create a focus for research activity and training for the next generation. A feature of the Centre will be its operating model. It will focus on fostering a strong interaction between industry - where many of the hard problems are brought into sharpest focus - and academia, which has a wide range of new mathematical and computational techniques that can be applied.