By Matthew Dublin

The National Nuclear Security Administration (NNSA) has a video describing some of the many non-nuclear research projects that are taking advantage of the NNSA's supercomputers, including the Roadrunner supercomputer located at the Los Alamos National Laboratory. While NNSA's IT staff in its Advanced Simulation & Computing division prepares its large-compute resources for the nuclear stockpile stewardship program, they regularly lend out computing hours to a variety of research projects.

In order to test Roadrunner and establish "system stabilization," LANL system technicians invited researchers working in a range of areas to run their applications and large data sets on the system. One project, led by researcher Bette Korber, a professor at LANL, used the NNSA supercomputer to model HIV proteins in an ongoing effort to find a possible cure for AIDS. Roadrunner provided Korber with insight into how the HIV virus replicates and aided her in the identification of common themes in all the variants that could one day be used to pinpoint a possible vaccine.

By Matthew Dublin

It looks like Giles Day knew whereof he spoke at last week's Bio IT World Expo when he called for a bit of "cloud sobriety," asking attendees: what if Amazon's cloud fails? What then? While most in the audience probably thought of that as the remotest of possibilities, like the entire national power grid failing, the unthinkable did in fact happen today. Early this morning at 1:48 AM PDT, Amazon's cloud failed, crippling many social networking sites including Foursquare, Quora,Reddit, and Hootsuite, Discovr, Wildfir, Livefyre, CampgroundManager, Totango, ESchedule, ZeHosting, Recorded Future, PercentMobile, the Cydia Store, and whatever other jobs were being run by private users at the time.

The technical failures affect Amazon EC2, Relational Database Service (RDS), Elastic Beastalk, CloudFormation, and Elastic Block Store (EBS).

In what is a sobering blow to the idea that the cloud's ubiquity and robustness insulates users from having the rug pulled out from underneath them, these failures were actually localized to Amazon cloud servers in Northern Virgina according to the AWS Service Health Dashboard. While technicians did maintain a detailed log of their efforts to get the servers back on line, it remains unclear as to the cause of the failures. The logs do seem to indicate that there were massive latency and error rates with EBS volumes and connectivity errors affecting EC2 instances, as well as other API errors. Let's just hope nothing more mission critical than a few Tweets and news about Lady Gaga was in jeopardy this morning. What was that about the cloud being ready for health care data?...

At a recent IT security conference held in New York City last week called The Computer Forensics Show, several security consultants balked at the idea of hosting health care data on the cloud, let alone sensitive business information or financial records. However, cloud users were encouraged by several speakers to take it upon themselves to audit the security of their cloud providers. The Cloud Security Alliance (CSA), a non-profit organization focused on promoting best practices for providing security assurance within cloud computing, was endorsed as a go-to resource for users looking to audit their cloud providers by John Kinsella, founder of Protected Industries, who is a co-chair of CSA working group. The CSA will soon be releasing guidelines to arm user with information on how to best kick the tires on their cloud and the right questions to ask providers.

By Matthew Dublin

One of the final talks at Bio-IT World Expo on Thursday was given by Giles Day, managing director of Distributed Bio, an informatics consultancy that caters to pharma and biotech companies. Day said that they typically sell their services to small companies with an informatics staff of usually no more than two or three people with small IT budgets and limited facilities. Most of these outfits are also managing increasingly complex automation in their workflows with unwieldy applications that produce exponentially expanding datasets.

Interestingly, Day said that a large part of his time is spent weaning clients off of their local compute clusters even after they have essentially hit the wall in terms of storage and compute power. Alas, the life of a cloud salesmen is not an easy one; the biggest barrier that Day and his company must help customers overcome in adopting the cloud is moving beyond their phobia of sending work outside the firewall and into the world beyond, or more specifically, up into Amazon’s EC2 cloud. Potential cloud users have a hard time believing that their data and intellectual property cloud ever be truly secure on Amazon’s EC2. But as he points out, roughly 98.9 percent of cloud users in the life science use EC2, and at the end of the day, Amazon really does know how to protect data in regulated environments as they handle tons of financial transactions every day without any breaches. Because of this, he argued, they have the security know-how that makes them one million times more secure than a pharma or biotech IT infrastructure could ever hope to be.

Security aside, the biggest issue with cloud computing has always been, and still is, I/O latency. There are several tools for addressing data transfer including rsync, Aspera, bbcp, and Bulk Ingest. But the folks running the Amazon cloud suggested to Day that the best method for transfer of large data sets is an application called Tsunami. Developed by researchers at Indiana University in 2002, Tsunami uses TCP (transmission control protocol) control and UDP (user datagram protocol) data for transfer over very high speed networks that are connected at long distance so that more throughput than is traditionally possible is achieved over the Internet.

One customer use case Day highlighted describes how the cloud can improve genomic annotation workflows, which is a classic embarrassingly parallel problem. Before the cloud, the client was running their genomic analysis pipeline with a 100 CPU cluster housed onsite, regularly processing upwards of 700,000 small genomes using a range of applications including Blast, hmmalign, hmmpfam, psort, and signalp, that resulted in terabytes of data. With their local cluster these jobs usually took two weeks to complete. But after coming to terms with the fact that they needed to rethink their whole approach, the client relented and switched over to the cloud, enabling them to reduce time-to-completion to just a few days.

Day did stop the singing the cloud’s praises long enough for a moment of “cloud sobriety” during which he pointed out that, because Amazon is really the only game in town for cheap and reliable cloud computing, and the one that the entire life sciences community interested in cloud computing is gravitating towards and developing methods for, what if EC2 goes out of business? While it’s hard to imagine the behemoth that is Amazon closing down its cloud operations any time soon, the question underscores the fact that this is still such a nascent technology, and when combined with the I/O issues and the learning curve to make workflows function smoothly with the cloud, IT staff need to proceed with caution and do the numbers before they commit.

By Matthew Dublin

Some of the presenters at a pre-conference cloud computing workshop held at the Bio-IT World Expo in Boston this week were not unlike an overzealous dog owner introducing their Great Dane to someone with a bad case of cynophobia. The general tone of many of the talks seemed to indicate that, for the most part, cloud computing is still in the get-to-know-me phase. Several of the speakers felt the need to play down the hype and instead play up the idea that cloud computing can provide a surprisingly user-friendly solution with lots of possibilities if only everyone just got to know it a little bit better. But they also understood that the real issue was not drumming up interest among researchers and IT managers, but rather, coming up with solid use cases that those folks can use when trying to convince their superiors or funding bean counters that cloud computing is not a technology to be feared and could potentially save money while enhancing research.

Case in point is the widespread use of REDCap — Research Electronic Data Capture) — a secure Web application for building online surveys and databases that has been implemented in the cloud at over 200 institutes to capture clinical research data. According to one presenter, Neil Bahroos, director of the Initiative in Biomedical Informatics at the University of Chicago, when coupled with the cloud, REDCap can be used in a mobile form so that physicians can feed data into a clinical study without ever having to worry about common security concerns, such as a misplaced laptop or hard drive, or just the accidental deletion of valuable data.

There was also much talk about Galaxy, which was described as a sort of "desktop informatics solution for next-generation sequencing" that when used with Amazon's EC2 is a good example of how the cloud be used to enrich a workflow. Users can present reproducible results through Galaxy's History feature, which allows collaborators to track the steps they took in an analysis and share them. In what started to feel like a late-night infomercial, the cloud was also pushed to attendees as something that no researcher should be without, especially if they have to pick up and move somewhere ("The cloud moves with you! No fuss, no muss!"), or for the those still on the fence, it was promised that more solutions like BioNimbus, a 212 node, 1568 core cloud that comes preconfigured with 2PB of data ready for you to use ("More data than you can shake a dongle at!"), will be coming online down the road; all users need to do is mount it onto their virtual machines and they'll be publishing highly accessed papers and winning grants in no time.

As far as identity and credential management in the cloud, there is a growing list of solutions that one could implement to ensure that only authorized users are permitted to gain access, including Openld (which is the same secure login backbone that Gmail uses), InCommon, Shibboleth, and MyProxy. Bahroos acknowledged that the rise in "private clouds" built by those tasked with shepherding and maintaining clinical data has been a natural result of the need to keep bosses happy by steering clear of breaking any HIPAA privacy rules. However, he hopes that this variant implementation of the technology, which seems to take away many of the benefits of what it's ideally supposed to provide (i.e. eliminating the burden of hardware), will soon die off as people become more comfortable the cloud; something that very well might happen if it doesn't jump up and bite anyone too badly.

By Matthew Dublin

Researchers at the University of Texas MD Anderson Cancer Center have their sights set on building the largest computing resource in the world dedicated solely to cancer research. A team lead by MD Anderson vice president and CIO Lynn Vogel has already begun implementing a system that uses a private cloud to provide computational power from 8,000 processors and hundreds of terabytes of storage along with a service oriented architecture-based Web portal called ResearchStation.

The roll out of the cloud coincides with MD Anderson's expansion into a third data center slated to come online later this summer, making it the second new data center opened at the center in a four-year period.

While Vogel says they initially looked at public cloud offerings, they opted for a private cloud solution not only for better performance, where there's less network latency than with a public cloud which is accessed through the web, but also because much of the data is sensitive patient information which might be more easily compromised off site.

"We've looked into this, but quite honestly, we've found on performance, access and in the management of that data, going to a public cloud is more risky than we're willing to entertain," Vogel says. "This goes directly to the point that this is identifiable patient data...and we're just not comfortable with the cloud given the actionable capability of a patient should there be a breach."

MD Anderson cancer researchers can request large blocks of the campus cloud on demand for analyzing human genomes and investigating radiation physics, epidemiology, and simulation for clinical trial activities.

"When you're in the business of biology, which we are, it's a different ballgame in terms of understanding the structures of data, the kinds of access and models used, and the applications that need to be available," Vogel says. "As much as public cloud providers would like us all to believe, this is not just about dumping data into a big bucket and letting somebody else manage it."

By Matthew Dublin

A team from the University of Illinois has published research in PLoS Computational Biology that details their use of molecular dynamics and steered molecular dynamics on GPUs to study point mutation induced Tamiflu-resistance in both avian and swine flu N1-subtype neuraminidases.

According the authors, Tamiflu is currently the frontline antiviral drug employed to fight the flu virus in infected individuals by inhibiting neuraminidase but drug resistance has become a critical problem due to rapid mutation.

The team, led by University of Illinois professor Zaida Luthey-Schulten, used the GPUs to construct an in silico model of the inner workings of a bacterial cell developed in collaboration with researchers at the Max Planck Institute of Biology in Germany and theoretical scientists at the University of Illinois.

Their simulations revealed "an electrostatic binding funnel that plays a key role in directing Tamiflu into and out of its binding site on N1 neuraminidase. The binding pathway for oseltamivir suggests how mutations disrupt drug binding and how new drugs may circumvent the resistance mechanisms."