Big scientific problems are often worked on by small teams of people using lots of machines. But those teams typically don’t include IT experts to help wrangle those machines. A group of molecular biologists can have access to the fastest compute cluster in town, but it won’t do much good without the know-how or time to manage it. The easiest solution would be to hire an IT person, but, you know, budgets and all …
Software that makes a cluster or grid accessible without an IT person might be the next best thing, which is the idea behind eXludus Technologies‘ new SimpliGrid program. The software runs under Windows XP or Vista and uses drag-and-drop operations to set up a grid and submit jobs to it. A demo made it look so simple even this reporter probably could do it. Certainly, that caveman on TV would have no problem.
Once SimpliGrid creates “MyCluster,” the user scientist drags systems from the Network Neighborhood to the MyCluster window. This is what it takes to provision a grid, eXludus says. Putting it to work doesn’t look like rocket science either (so to speak). The user just drags jobs from whatever window they’re in to the cluster window. “Without any other intervention from the user, the processing jobs will be distributed out to systems in the grid,” says Dale Geldart, executive vice president of sales and marketing at eXludus.
“Clusters and grids have become the standard for meeting higher-performance requirements, as they provide the means to aggregate compute power,” Geldart says. “The challenge for end users, though, is that it generally takes specialized skills and knowledge of complex workload management tools to provision and manage a cluster or grid, leaving the possible use of such architectures out of the reach of those who work outside IT. With SimpliGrid, we’re trying to open up grid benefits to researchers, scientists, and teachers who have little system management experience.”
Example? “A bioscience researcher working in a medium-size lab — say, 25 to 50 systems, many of which may be desktop systems for a group of users — could perform gene sequencing/analysis much more efficiently than would be otherwise possible.”
Geldart says departmental labs of “many flavors” also would be beneficiaries, as well as smaller investment companies running option pricing models, but for now the developer is focused on life science applications. “Too often we see researchers who have all these systems in the lab but no ability to harness this compute power short of manually distributing work” machine by machine, he says.
Getting Grids to Those Who Need Them
Complexity deters many scientists from tapping into the compute power that’s very often just sitting there. It’s like learning how to master Excel, but on a cosmic scale: you know you could do a lot with it, but would prefer someone else do it for you.
“For an average researcher working in a smaller lab, the big challenge is grid implementation complexity coupled with lack of available budget to hire IT administrator types. This type of user probably doesn’t even think ‘grid’ because of an immediate assumption that grid management and use is beyond their capability. These people don’t want to become IT admins. So they end up with a lot of aggregate processing capacity scattered across the lab but no easy means to tap into it.”
J.W. Bizzaro would know about that. As founder and chairman of The Bioinformatics Organization, he’s a champion of “bioinformatic practitioners” and computational scientists everywhere. (eXludus has given organization members the opportunity to beta test the software and license it for $99 when it becomes available later this year.)
“One of the attractive things about the platform is its ability to create sort of an ad hoc grid of Windows systems in the lab,” Bizzaro says. “I’ve seen other attempts at creating these kinds of grids, but they required rewriting programs or wrapping them in a certain language. SimpliGrid does not.”
Ease of use is the thing, Bizzaro says. “SimpliGrid picks up the batch file and distributes it to the systems in the grid. Then it balances the workload, tries to optimize the way the job is split up in order to finish in optimal time, then returns the data as a file in the same folder. It’s very nice in that it’s very straightforward.”
SimpliGrid should be ready for commercial release in August, the company says.
eXludus develops software for managing, virtualizing, and provisioning datacenters. Products include the Optimizer and Replicator suites for improving grid and cluster performance. In February, SGI announced that it would embed eXludus’ Grid Optimizer software in its BioCluster system.