By Matthew Dublin

Oak Ridge National Laboratory (ORNL) is partnering with Cray to build a 20 petaflop system that will be dubbed “Titan.” According to a presentation by project leader Buddy Bland, at the Oak Ridge Leadership Computing Facility in the Tennessee, assembly on the system will start this year and is expected to be running at its full 20 petaflop capacity by next year.

The estimated cost for the giant machine is somewhere in the neighborhood of $100M and will be made up of Cray’s new “Gemini” XE interconnects and a new memory system the system’s team is called “globally addressable memory”- more information on this to follow. Titan will also sport 16-core AMD processors and GPUs as co-processors, an increasingly common feature of new HPC systems over the last two years, and will run an updated version of Cray’s Linux as an operating system.

By Matthew Dublin

Researchers at the University of Manchester now have an easier, desktop drag-and-drop method by which they can take advantage of the university’s Condor pool. Called DropAndCompute, this application lets users submit and manage large data analysis jobs on a grid or cloud without needing any expertise with the system’s operating system or Unix or Linux command line tools.

DropAndCompute was initially inspired by and built on top of Dropbox, the freely available graphical file management application.

The University of Manchestor's Ian Cottam, one of the lead developers on the project, has a blog post describing how the solution evolved from a homegrown job management application built on top of Dropbox into a standalone application.

Here is a video of the first version of DropAndCompute that was used on a Condor pool within the university’s interdisciplinary biocentre:

The latest version of DropAndCompute no longer relies on Dropbox and is a standalone “local version” that only requires users to mount a folder on the submit node using their local computer and the same drag-and-drop approach in order to initiate jobs to run on their institute’s HPC resources:

By Matthew Dublin

Genomatix has announced this week the release of mygenomatix, a “cloud-like” computing service that incorporates its in-house next generation sequencing data analysis platform. The new offering claims to provide users with analyzed data in a “matter of one to two weeks” and says the security is much better than “standard cloud models.”

The offering is slated for launch this April and from CEO Marty Seifert’s description in the initial announcement, it sounds like what they have basically created is a service model where the work around for security holes is…snail mail.

“We have been exploring a cloud like model for quite some time, and our service model addresses the issue of security by getting the data to our computers via a hard disk shipment program,” stated Seifert in the press release. "With mygenomatix, anyone doing NGS data analysis now has access to an easy entry path to our technology and databases."

Who would have thought all cloud computing was really lacking was a postage machine? At some point in the future maybe a consensus will be reached on exactly how to define the "cloud" because more often than not it seems to be inserted in marketing copy in place of less sexier concepts such as plain old software as service or the ancient utility computing model.

Other vendors are not willing to give up on the idea that the cloud can be secure enough in its own right. SecludIT has released a new tool that capitalizes on their “Elastic Security technology” which allows users to dynamically adapt security settings as their cloud computing infrastructures change. SecludIT’s Amazon Detector for Amazon EC2 is an automated security even detection tool that is designed to help users isolate holes on security groups and applications. The company claims to employ a new paradigm called “auto checks” that lets users set the correct security checks and alerts automatically.

In other cloud news this week, Complete Genomics announced on Monday that it will now offer its customer the option of storing and visualizing their sequencing data with DNAnexus’ cloud computing platform. Complete Genomics’ customers will have access to a informatics tools through the DNAnexus cloud, built on Amazon’s EC2, where they will be able to visualize and query multiple complete human genomes. DNAnexus has also announced that in the near future they will integrate Complete Genomics analysis tools functionality, an open source software project for downstream analysis of data produced by Complete Genomics.

By Matthew Dublin

A new study by researchers at the University of Warwick’s Institute for Digital Healthcare and the Centre for Health Informatics and Multiprofessional Education at UCL Medical School says that open source software may be more secure than closed source or proprietary software. The impetus for the study is based on the observation the well documented troubles the health care industry has with vast costs, lack of fault tolerance, and incompatibility of systems preventing database architectures and web applications from talking to each other.

The long-running debate is buoyed up by many a salient point on both sides, so the results of this study will hardly close the book. However the authors maintain that open source software is a secure and cost-effective prescription to cure the health care industry’s IT woes. Their paper refutes critics’ claims that because the source for code is available to the general public the software is inherently more vulnerable for exploitation and attack.

"Proprietary systems often rely on a 'security through obscurity' argument, that is systems that hide their inner workings from potential attackers are more secure. However security through obscurity alone completely fails when code is disclosed or otherwise discovered using tools such as debuggers or dissemblers,” says Jeremy Wyatt, a professor at the University of Warwick's Institute for Digital Healthcare."Worse, it has been suggested that the cloak of obscurity tends to encourage poor-quality code. Opening the source allows independent assessment of the security of a system, makes bug patching easier and more likely, and forces developers to spend more effort on the quality of their code."

The researchers' paper entitled "Open Source, Open Standards, and Health Care Information Systems" has been published in the Journal of Medical Internet Research.

By Matthew Dublin

The Swiss National Supercomputing Centre in Manno, Switzerland, has announced that it will install a next-generation Cray XMT supercomputer to accelerate a range of research projects including the HP2C platform, an initiative that supports several efforts focused on developing programming and database design approaches to take advantage of the generation of supercomputers expected to become available by 2013. These projects include the massive "Selectome" database that stores information on genes which have evolved under Darwinian selection usually thought to be involved in complex diseases.

CSCS, a long-time Cray site, currently has a Cray XT5 supercomputer called "Rosa" and was also the first recipient of the Cray XE6 system. The site will use its next-generation Cray XMT supercomputer for solving problems that require large-scale data analysis and will become part of a new project called 'Eureka.' The proposed facility will be used for large-scale analysis of unstructured data and data mining, and is designed for parallel applications that are dynamically changing, require random access to shared memory and typically do not run well on conventional systems.

Each processor in the Cray XMT system can handle up to 128 concurrent threads and is designed to scale from 16 processors up to multiple thousands of processors that can operate on multiple terabytes of shared physical memory.

By Matthew Dublin

A tool offered by TeraGrid aims to help users easily navigate the multiple systems the make up the geographically distributed grid resource that is comprised of a very diverse hodgepodge of HPC hardware. TeraGrid's new Common User Environment simplifies access by identifying commonalities across systems and eliminating many of the differences.

"CUE was the solution to a number of nagging problems from an EOT (education, outreach, and training) point of view," says Jeff Pummill, a TeraGrid Campus Champion Leadership Team Member based at the University of Arkansas-Fayetteville. "New TeraGrid users want to test-drive several resources. Without CUE to unify the machine environments, they can become discouraged spending more time learning about each machine's unique features than on their research."

CUE consists of three major components:

Management System: Using provided examples of core modules that have been developed for all TeraGrid production systems, users can customize a command-line interface for their unique needs as well as a default interface that features the same naming conventions for basic functions on multiple resources.

Variables Collection: Allows users to take advantage of common environment variables and parameters in the global ssh_config file with values that are defined across all platforms. This includes a short-hand method of configuring TeraGrid host aliases, which eliminates the need to enter fully qualified login host names when using ssh or gsissh, as well as a standardized shell prompt that indicates which resource an open shell window is connected to.

Testing Platform: Example programs that can be compiled and executed through the CUE to demonstrate its use on all TeraGrid machines allows users to make valuable comparisons. This allows users to learn how to write, compile, and run fully functional programs by following examples of simple scientific MPI code written in C and Fortran programming languages.