Bristol-Myers Squibb has agreed to evaluate Entropia’s desktop PC-based distributed computing technology. Entropia said the deal represents a prime opportunity to demonstrate the potential of an approach that pharmaceutical and biotechnology companies have been slow to adopt.
Martin Stuart, vice president of life science sales at Entropia, said the company has a number of similar pilot projects at other, undisclosed, pharmaceutical companies. He is optimistic that several of these firms are ready to emerge from the proof-of-concept phase and will soon adopt the technology.
“The pilot demonstrates to the organization that, yes, the technology can harvest these spare PC cycles, yes they can be directed toward computationally intensive tasks, and yes, the company does have a huge virtual supercomputer sitting within its organization,” said Stuart.
“I found the whole economic model behind the technology very compelling,” said Richard Vissa, executive director of global core technologies, informatics, at Bristol-Myers Squibb’s Pharmaceutical Research Institute. “PC vendors are now shipping gigahertz machines at the entry point. So the price-performance curve of PC computing hardware has been increasing much steeper than traditional servers.”
Entropia estimates that a typical PC remains idle as much as 95 percent of the time, wasting available processing cycles. Stuart said that as few as 2,500 processors with an average clock speed of about 500 MHz would supply a company with around a teraflop of compute power. “And when you consider that many of these organizations have tens of thousands of machines, it becomes a very compelling technology for them to look at,” he added.
The typical cost savings is dependent upon the applications an organization runs in the environment, Stuart said, but noted that because firms have already made the initial investment in their PC network, the ability to increase the utilization of the existing machines is “an attractive proposition.”
Vissa said that several hundred BMS PCs were involved in the pilot study. In addition to evaluating the system’s performance, throughput, and operating costs relative to a comparable supercomputer, he said his team would need to judge how well it scales when additional PCs are added, as well as whether the software runs unobtrusively on users’ desktops. If BMS does decide to implement the technology, Vissa said, it would most likely complement, rather than replace, a supercomputer.
Entropia is not alone in realizing the potential value of this approach to pharmaceutical companies. Platform Computing, which has until now focused its efforts on load distribution software for compute clusters, just released its first product to harness unused processing power from desktop PCs. Platform launched the software, LSF ActiveCluster, at the Beyond Genome conference in San Francisco two weeks ago, in a move to target the life sciences market.
With a current customer list that includes Boehringer-Ingelheim, Celera Genomics, Lion Bioscience, Merck, Pfizer, the Sanger Center, and the Whitehead Institute — ”several” of whom are beta-testing the new LSF ActiveCluster software, according to a spokeswoman — Platform may give Entropia some stiff competition in this space.
In addition, other companies, such as Juno, Parabon, and United Devices, provide a similar service to companies by utilizing spare PC power over the Internet. Juno signed its first life science customer for this service in May, when Pittsburgh-based bioinformatics incubator LaunchCyte agreed to use the Juno system to increase the development speed of its projects.
But despite the growing competition, Stuart said he’s optimistic that some of Entropia’s pilot projects in the pharmaceutical sector will convert to sales before the end of the year. “I’m delighted to say that the pull from the [life science] marketplace has just been astonishing,” he said. “We’re very busy here at this point in time, and it’s going to get busier.”