Mouse Informatics at the Jackson Lab

The Jackson Lab's Carol Bult discusses her institute's mouse informatics resources.

A Conversation with Bud Mishra

We sat down with NYU's Bud Mishra to discuss ongoing project in the NYU Bioinformatics lab and more.

A Conversation with Eric Schadt Part II

Here's part two of our conversation with Eric Schadt.

A Conversation with Eric Schadt Part I

Eric Schadt discusses the work of Mount Sinai's Institute for Genomics and Multiscale Biology.

A Conversation with Andrea Califano

In this video, Columbia University's Andrea Califano discusses algorithm development and more.

Japanese Supercomputer Reaches 10-Petaflop Barrier

By Matthew Dublin

Japan's Riken Institute has pushed the K Computer — currently the world's fastest supercomputer — to the 10 petaflop barrier or 10.51 quadrillion floating point operations per second.

The K Computer is comprised of 864 server racks and over 88,000 interconnected CPUs, Fujitsu's SPARC64 VIIIfx processors. According to the industry Linpack benchmark, the K Computer’s average performance is about 93 percent of its theoretical peak speed of 11.28 petaflops.

Back in June, the K Computer topped the Top 500 list with a peak performance of 8.162 petaflops, however IBM and Cray are hot on its tail with 20 petaflop machines in the works for the Department of Energy that are slated to go live at the end of the year.

The K computer has 640,000 cores working around the clock on complex mathematical problems and eats about roughly 10 megawatts of energy, which is surprisingly energy efficient when compared to other supercomputers around the world.

The K Computer cost over 100 billion Yen — $1.25 billion US — to design and build.

Organic Parallel Processors

By Matthew Dublin

A team of Japanese researchers at the National Institute for Materials Science in Tsukuba have designed a parallel processor that mirrors neurons in the human brain using organic molecules.

The organic processor uses a ring-like molecule called 2,3-dichloro-5,6-dicyano-p-benzoquinone, or DDQ.

The circular DDQ molecule can exist in four different conducting states and is from one state to the other by hitting it with varying voltages using the tip of a scanning tunneling microscope. By placing two or more molecules next to each other they can be connected — the team was able to connect 300 DDQ molecules as a single layer on a gold substrate, making one massively parallel computing infrastructure with only a single layer of organic molecules.

Various concepts of a wireless molecular circuit: In panel a) you see the DDQ molecule that the researchers used. The next panels show the molecule in its various states and configured as organic circuits.

Anirban Bandyopadhyay, a scientist from the National Institute for Materials Science, says that “generalization of this principle would change the existing concept of static circuit-based electronics and open up a new vista of emergent computing.”

China's Newest Supercomputer Boasts Homegrown Chips & Advanced Water Cooling

By Matthew Dublin

China's newest supercomputer, the Sunway BlueLight MMP, has caused quite a stir in the international HPC community this week. Much to everyone's surprise, the Sunway system is performing about 1,000 trillion calculations per second using 8,700 Chinese-made microprocessors.

Last year, China made waves with its Tianhe-1A supercomputer, which briefly occupied the number one ranking on the Top500 list as the world's fastest supercomputer. Although Tianhe-1a did sport an advanced interconnect system designed by a Chinese team, the actual processors chips were made by Intel and Nvidia — which allowed HPC developers and researchers to breath a sigh of relief. So considering that China was thought to be about three generations behind chips makers in the United States, Japan, and Taiwan, the community is keenly interested to take a closer look at the Sunway's ShenWei SW1600 microprocessors.

But more interesting perhaps is that fact that the Sunway uses only one megawatt of power and yet is about 74 percent as fast as Oak Ridge National Laboratory's Jaugar supercomputer – currently the third fastest machine in the world – which consumes . Also of note is Sunway's advanced water-cooling system, which could not only help keep power costs down but could also pave the way forward for an exascale-capable system in the future.

“Getting this cooling technology correct is very, very difficult,” says Steven Wallach, chief scientist at Convey Computer, in an interview in the New York Times. “This tells me that this is a serious design. This cooling technology could scale to exaflop. They are in the hunt to win.”

Sunway's liquid cooling components:

The Sunway's ShenWei SW1600 microprocessors:

Some Sunway stats:

▪ Installed in 09/2011 at NSC in Jinan, Shandong,China

▪ Petaflop, number 2 in top100 2011 in china, peak 1.07016 PFlops, Sustained 795.9 Pflops, linpack efficent 74.4 percent

▪ 8700 ShenWei SW1600 CPU, designed in chinese, manufactured in Shanghai

▪ power 1074KW, water cooling

Cornell Launches Subscription Cloud Service

By Matthew Dublin

The Cornell University Center for Advanced Computing (CAC) has launched an on-demand research computing cloud called "Red Cloud". The cloud is available to both Cornell University researchers and other academic institutions.

There are currently two available services, "Red Cloud" and "Red Cloud with MATLAB."

"Red Cloud" is an Infrastructure as a Service (IaaS) that runs Eucalyptus, open source cloud computing platform, that will provide subscribers with root access to virtual servers and virtual disks.

“Red Cloud with MATLAB,” is a Software as a Service (SaaS) that runs MATLAB Distributed Computing Server and features GPUs. Users can develop their applications using the Parallel Computing Toolbox and then scale up to the cloud with the MATLAB Distributed Computing Server.
 
The Red Cloud sports an 10Gb Ethernet network for moving large datasets in and out of the cloud with 50GB of available storage per user.  

"Red Cloud" subscriptions for Cornell staff is $500 for 8,585 core hours and $750 for "Red Cloud with MATLAB." For non-Cornell researchers, subscription prices run $750 and $1,200 respectively.

“Cloud computing has the potential to advance U.S. research discoveries and better prepare students for STEM careers by making computing, software, and data resources more readily available at an economy of scale,” says CAC director David Lifka.

Amazon Cloud Security Gaps Exposed

By Matthew Dublin

You better pause your data uploads: researchers at the Ruhr-University Bochum announced today that they have identified some serious security gaps at Amazon Cloud Services.

The team was able to attack AWS clouds using XML signature wrapping and cross site scripting. There are several types of XML Signature Wrapping attacks which involve altering the signed message a web service client sends to the receiving web service. The attacker is able to change the content of the signed message without invalidating the signature which allowed the researchers to take over administrative control of a cloud, including creating new instances as well as adding or deleting virtual compute images.

Cross site scripting attacks also allowed them to get into the Amazon shop itself where they could access customer data, including authentication data and text passwords. This type of attack enables hackers to inject client-side script into web pages in order to bypass access controls.

Once the researchers completed their attacks, Amazon was notified about the security gaps and proceeded to shut them immediately.

"Based on our research results, Amazon confirmed the security gaps and closed them immediately...A major challenge for cloud providers is ensuring the absolute security of the data entrusted to them, which should only be accessible by the clients themselves," says Jörg Schwenk, who headed up the AWS attacks.

Schwenk is also quick to point out that private clouds, which are typically hosted at a site internally versus a public provider like AWS, are not safe from these types of attacks either. Eucalyptus, an open source project widely used to implement Cloud solutions within companies, did expose the same weaknesses. "A rough classification of cloud technologies cannot replace a thorough security investigation," he says.

There's a Processor in Your Guts

By Matthew Dublin

Researchers at the Centre for Synthetic Biology and Innovation and the Department of Bioengineering at the Imperial College London may have pushed biological computing closer to becoming a reality one day.

In a paper published today in Nature Communications, the team described how they built some advanced logic gates constructed with just gut bacteria and DNA.

Using E. Coli that was altered with modified DNA, they constructed a type of logic gate called an "AND Gate" that performed the same "on" and "off" process as its electronic equivalent when stimulated by chemicals. In another experiment, the researchers showed that they could create a "NOT gate" and combined it with the "AND gate" to produce a "NAND gate."

While computer processors that process bits of information with molecules as reliably or powerfully as the processors we have currently, the UK team's research does show some serious advancements in the field.

Previous studies has been published showing that biological logic gates could be made, but the team claims that their gut and DNA logic gates behave more like their electronic counterparts. And in a triumph of nano engineering, these biological gates are also modular, meaning that they can be fitted to make different types of logic gates, possibly even more complex biological processors down the road.

Need More Data Storage? Just Add Salt

By Matthew Dublin

Researchers at Signapore’s Institute of Material Research and Engineering (IMRE) and National University of Singapore are using table salt to increase the data recording density of hard disks.

While conventional hard disks use randomly distributed nanoscopic magnetic grains to store up 0.5 terabits of data, the research team used nanopatterning to tightly pack the arrays of magnetic bits to store up to 3.3 terabits/in2 of information.

The team discovered that by adding sodium chloride to a developer solution used in existing high-resolution e-beam lithography processes used to create super fine nano-sized structures, they were able to produce highly defined nanostructures down to 4.5 nanometers half pitch, without the need for expensive equipment upgrades.

Comparison of different bit densities:

What this could mean down the road is that hard disk drives currently capable of storing 1 Terabyte (of data today could, could eventually hold 6 TB of information in the same form factor.

“What we have shown is that bits can be patterned more densely together by reducing the number of processing steps”, says Joel Yang, the IMRE scientist who heads the project. “In addition to making the bits, we demonstrated that they can be used to store data.”

This ‘salty developer solution’ approach was actually invented by Yang when he was a graduate student at the Massachusetts Institute of Technology.