This article has been updated to correct previously reported details about JAX existing compute infrastructure including details about the size of the cluster and its composition.
Convey Computer said this week that the Jackson Laboratory has purchased one of its new hybrid-core computing systems to handle its next-generation sequencing analysis needs.
The company officially launched the new line of hybrid-core computers — the HC-2 series, which is comprised of the HC-2 and HC-2ex — at the Bio-IT World conference in Boston this week.
The HC-2 systems offer the same combination of field-programmable gate array and x86 chips as the company's previous HC-1 line, but Convey claims that they beat their predecessor systems in terms of performance, functionality, and energy efficiency. Specifically, the company said that the HC-2 systems increase application performance two to three times over the HC-1 "several orders of magnitude" over commodity servers.
The starting price for the HC-2 system is $40,000 and the pricing changes based on the customizations made to the infrastructure as required by the end user, George Vacek, Convey’s director of life sciences, told BioInform.
The HC-2 system implemented at JAX includes 224 gigabytes of memory, 6 x 600 GB 10K RPM disk drives, and gigabit ethernet cluster interconnect, Convey told BioInform.
According to hardware vendor, JAX is using the system in its next-gen sequencing data analysis projects.
JAX conducts a wide range of basic and applied research on topics such as cancer, aging, cardiovascular disease, immunology, reproductive biology, and neurobiology. JAX uses high-throughput sequencing for a variety of purposes, including the discovery of spontaneous mutations, strain-specific variation, and genome-wide analysis of gene expression.
The new system’s capabilities will enable JAX researchers to perform whole-genome studies that were “impractical” on the lab’s previous compute infrastructure — a cluster of several 32-core servers, Convey said.
For example, the lab is using Convey’s system, which includes the firm’s implementation of the Burrows-Wheeler Aligner, as part of its efforts to identify disease-causing mutations in the mouse genome. The lab also plans to use the system in projects that require de novo assembly.
“Once we could afford whole-genome sequencing, we found a significant bottleneck in the time required to process the data,” Laura Reinholdt, a research scientist at JAX, said in a statement. “That’s when biologists here began to seek tools and infrastructure to more expediently manage and process the expanding volumes of NGS data.”
Glen Beane, the senior software engineer at JAX, told BioInform that the laboratory looked at “various options for hybrid systems” before settling on Convey.
“We found GPUs weren’t a good fit for alignment. There are packages that do alignment, but the performance isn’t that compelling,” he said. “We looked at other FPGA system vendors, but they didn’t have the number of tools Convey does or the system wasn’t as easy to use.”
Convey’s hybrid-core architecture is made up of Intel processors with a coprocessor comprised of FPGAs. In addition to BWA, the company offers a package that speeds algorithms for short-read assembly (BI 5/20/2011) and an implementation of the Smith-Waterman algorithm (BI 11/20/2009).
It provided the best combination of power consumption, space, and performance for a “fixed among of dollars,” Beane said.
He added that “a developer community is evolving around the Convey systems where we could share third-party tools.”
Convey’s “open platform” model sets it apart from FPGA vendors like TimeLogic whose solutions don’t let users “see under the hood” and who don’t offer open source application and algorithms like BWA that are familiar to customers in the bioinformatics space, Vacek noted.
Additionally, Convey has designed its system to be easy to use and to function in the same way a standard compute environment works, he said.
“The feel of the system is just like anything the user would have with a [standard] computer. It’s just running faster,” he said.
The new system will complement JAX’s existing system, a cluster based on 8-core AMD Opteron processors with four processors and 128 gigabytes of memory per node.
“Rather than add five more nodes to our cluster, this system will essentially allow us to add one optimized alignment node that we can use instead,” he said.
There are two ways to look at the Convey addition, he continued. “One is we are scaling up because we need to add more alignment capacity … the other aspect is we looked at how scaling up could help do things that we weren’t able to do before. The Convey system also helps achieve that goal.”
Chuck Donnelly, director of computational sciences at JAX, said in a statement that “initial benchmarks” indicated “a ten-fold performance improvement in BWA” using Convey’s system and accelerated version of the algorithm compared to running the original algorithm on the lab’s existing infrastructure.