This week’s Bio-IT World Conference & Expo in Boston, Mass. gave Apple an opportunity to show off its recently introduced Xserve cluster configuration. What’s more, one vendor showed off a new application designed specifically for scientists who would like to add Xserve-based clustering to their network quickly and easily.
Apple’s booth was highlighted by a pair of rack systems populated by Xserves set up in cluster configurations, complete with Xserve RAID storage. A table offered passersby a look at Mac OS X running on a new Power Mac G4, iMac and a 17-inch PowerBook G4. The PowerBook predictably drew the most interest from the steady stream of showgoers who paid Apple’s booth a visit.
Apple was also featured in a Wednesday afternoon technology showcase in a theater area on one corner of the exhibit hall area on the second floor of Boston’s Hynes Convention Center. Doug Brooks, Apple product manager, server hardware, opened up the session by offering attendees a brief history of the Xserve.
Brooks noted that the Xserve, introduced less than a year ago, has opened up a new market for Apple. While the company has long had workgroup servers and other server products available for corporate, educational and engineering markets, the Xserve is the first 1U rack-mount configuration Apple has ever offered. Brooks also touched on the Xserve RAID, Apple’s complementary rack-mounted storage system with up to 2.52TB of storage capacity, and the newest member of the Xserve family, the Xserve cluster configuration — a special stripped-down Xserve model better suited for distributed computing applications.
Following Brooks’ presentation, Dr. William Van Etten of The BioTeam gave a presentation entitled Scalable Informatics for the Scientist. He painted in broad strokes the necessity and availability of scalable distributed computing environments for scientists and researchers.
Until now, said Van Etten, scientists have often had to spend weeks or even months getting a distributed resource management system up and running to their liking — he offered a veritable laundry list of networking protocols, file systems, services and other technical minutiae that have to be manipulated and synchronized in order for a distributed computing cluster to work effectively. The combination has often proved frustrating for researchers who would rather spend time doing research and IT personnel who would rather spend time administering computer systems, but it’s a big part of the bio-IT business as evidenced by a number of vendors on the show floor besides Apple offering distributed computing hardware and software solutions.
The BioTeam has been providing solutions in the distributed computing arena for a while for a client list that includes research teams at Harvard University, UC Berkeley and the Dana Farber Cancer Institute. The BioTeam decided to leverage Mac technology when Texas A&M University asked them to create a distributed computing cluster suitable for scientists without any Unix training.
The BioTeam ported Sun’s GridEngine — an open source distributed resource management application — to Mac OS X, and deployed it with Platform Computing’s Platform LSF, a distributed computing workload management system. The BioTeam also modified PISE — a toolkit developed by the Pasteur Institute’s Catherine Letondal, which provides a Web interface for molecular biology programs — to work with the Xserve cluster.
The resulting distributed computing system has been deployed on a 10-CPU 5-node Xserve cluster at Texas A&M with outstanding results, according to Van Etten. The BioTeam has been so pleased with the results that they’ve taken their work a step further by creating their first product, which they call iNquiry.
iNquiry combines the technology The BioTeam developed for Texas A&M into a system that other bioscientists can use to create their own Xserve clusters in a matter of minutes instead of weeks, according to Van Etten. The secret is another Apple product — the iPod.
Wholly self-contained in about 2GB of storage space on the iPod, iNquiry uses a Perl-based script that’s controlled through a simple graphical configuration utility. The user tells the configuration utility how to configure the Xserve cluster, how many nodes it has, how the network is configured, and how to use the individual drive bays in each Xserve.
The Xserve cluster’s head node is then rebooted off the iPod, through the Xserve’s FireWire interface. A script automatically executes and configures the Xserve, which takes about 15 minutes.
“Mac OS X makes this all so terribly easy,” said Van Etten.