The Red Storm has passed quietly over Celera’s radar screen. Sandia National Laboratories recently terminated a cooperative research and development agreement it signed in January 2001 with the company to build a petaflop-scale supercomputer. Compaq, “a major partner” but not an official signee of the CRADA, has also withdrawn from the project. Instead, the DOE lab has entered negotiations with Cray for a $90 million contract to build, install, and support the so-called Red Storm supercomputer.
Citing the dramatic business and organizational changes that have occurred over the past year at both Compaq and Celera as the basis for terminating the deal, Sandia’s Grant Heffelfinger noted that “the vision that was in the CRADA, and Sandia’s interest in that vision, is not diminished, nor has it changed. We’re simply not pursuing it with Celera.”
Celera spokesman Robert Bennett agreed that the company’s rapid evolution over the last year bumped the CRADA off its to-do list. “We felt we both got out of it what we could, and at this point our needs are really divergent beyond what that exchange covered,” he said.
Heffelfinger, deputy director for materials science and technology at Sandia and technical leader on the CRADA, said the project was initiated to develop the architecture, operating system, and other supporting technologies for Red Storm, a massively parallel system that would scale up to 10 teraflops by 2002, 100 teraflops by 2004, and eventually a petaflop. While the supercomputer was planned to support Sandia’s primary mission of national security research, the collaboration with Celera was seen as an opportunity to “bio-ize” the lab’s computing capabilities, Heffelfinger said. A life science component of Red Storm is still envisioned, he added, but a new partner has not yet been identified for that aspect of the project.
Heffelfinger didn’t rule out the possibility that former Celera president Craig Venter, who played a key role in the initial negotiations with Sandia, might resurface in the Red Storm picture. “We certainly have a historic relationship with Craig Venter, so that’s obviously a place to start,” he said. “It remains to be seen what [Venter’s new non-profit organizations] TCAG [TIGR Center for the Advancement of Genomics] and IBEA [Institute for Biological Energy Alternatives] decide their computing needs will be and how they want to go.”
A TCAG spokeswoman said it was too soon to comment on the organization’s plans.
In with a Bang, Out with a Whimper
The collaboration between Celera, Compaq, and Sandia was ushered in with typical pre-genome hoopla last January. Former DOE secretary Bill Richardson touted the “marriage of molecular biology with high-performance computing” as he presided over a special signing ceremony between Venter and Sandia president Paul Robinson.
But that was before Compaq announced its plans to phase out the Alpha chip, while its merger with Hewlett-Packard was still a twinkle in Carly Fiorina’s eye, and Celera was riding high on the promise of its data-centric business model.
With two-thirds of the project’s players barely recognizable a year and a half later, it seems inevitable that the collaboration came to an end.
According to Heffelfinger, following an initial burst of productivity in which Celera’s bioinformatics algorithms were optimized to run more efficiently on its existing Compaq compute farm, Celera’s contributions to the project fizzled out in the fall of 2001, and Sandia ended the partnership in an “amicable departure.”
“We just sort of parted company,” he said. “They’re going in a different direction than they were and we still want to go in the same direction.”
Bennett, who described the parting as a “wind-down,” said both parties benefited from the project. “Our plans are to independently take those learnings forward and apply them in our respective applications in what they do and what we do in biology.”
According to Sandia officials, there are no strings attached to ending a CRADA, the terms of which are mutually agreed upon by participating parties. Neither Celera nor Sandia contributed any IP to the project and Celera will not be penalized financially.
“The agreement itself is really around how you manage shared intellectual property, and there wasn’t any — it didn’t exist long enough for that to happen,” said Heffelfinger.
Cray Snags Contract from ‘Distracted’ Compaq
Meanwhile, on the computational side, the Red Storm project was far from Compaq’s top priority over the last year. Even before it modified its technological roadmap and merged with HP, Compaq’s involvement in Red Storm was essentially limited to its role as a mutual computational technology base for the two CRADA partners, who each boast large installed Compaq clusters. Compaq’s role remained undefined beyond the first stage of the project, which was to improve the productivity of Celera’s existing computing resources.
“[Compaq’s] involvement was most likely to be most heavily in the context of the hardware and the porting to the hardware of these new capabilities we were planning to develop in the context of Sandia and Celera’s follow-on procurements,” said Heffelfinger. But when Sandia issued a competitive call for proposals for the Red Storm procurement, Cray won out over a number of undisclosed competing vendors.
Celera’s future computing requirements are now under review within the context of a new set of computational challenges, including data analysis and management for proteomics, new docking algorithms for lead discovery, and an expanded knowledge business platform under Applied Biosystems. “As a consequence we are reevaluating our storage and data management requirements,” said Bennett.
(Genomes to) Life Goes on at Sandia
Heffelfinger noted that Sandia’s work with Celera — short-lived as it was — only served to whet the lab’s appetite for life sciences computing. In addition to a commitment to recruit a new biological partner for the Red Storm project, the lab is setting its sights on the computational infrastructure that will be put in place for the DOE’s Genomes to Life program. Heffelfinger is the principal investigator on a $6 million joint Sandia/Oak Ridge National Lab GTL proposal to build the computational capabilities to study carbon sequestration in the marine bacterium Synechococcus.
Heffelfinger said the lab’s experiences with Celera have spurred its efforts to play a larger role in the DOE’s biology research portfolio. “If I could go back in time to when we started the discussion around the [Celera] collaboration and know everything I know about how it all turned out, I would pursue it again in a heartbeat,” he said.