TIGR conducts computing-intensive research involving the structural, functional and comparative analysis of genomes and gene products in viruses, bacteria, archaea and eukaryotes.
TIGR recently announced that it has now entirely transitioned its IT operations from 15 HP Alpha servers onto three Sun Fire x64 servers to its complex genomic assembler for molecular research.
The director of IT at TIGR, Vadim Sapiro, was given several mandates for the project.
As a public enterprise, the Institute needed to make more of its software open-source so it is freely available to researchers.
Originally aligned with the company Celera Genomics, TIGR took software that was once proprietary to Celera and made it open source over the course of several years from about 2000 to 2004.
The new server system needed to be able to support the open-source tools that TIGR now provides.
"The open-source software was highly dependent on the alpha architecture," Sapiro said. "Because of that dependency, we had to go out and procure alpha servers to run the software."
Not only were the software requirements a priority, but cost and performance were key as well.
As a public entity funded solely by grants, TIGR was also highly cost conscious.
Given the enormous technological leaps since 1999, this system of course cost significantly less than the prior one to acquire.
A server comparable to one of the 15 that TIGR had been running cost about $400,000 to purchase and an additional 10 to 15 percent a year for support and maintenance costs, in addition to cooling and the form factor expenses.
By comparison, the three new servers were only about $30,000 to purchase and came with three years of free technical support.
Energy consumption for cooling was reduced by 70 percent and the form factor expenses, given the far smaller amount of space required, were slashed. (Sapiro said TIGR benefits from a nonprofit pricing structure.)
TIGR had an existing relationship with Sun, which Sapiro says was instrumental in the choice.
But Sun was also the only company that would commit to supporting the hardware for free, a promise that was quickly put to the test when the very first server Sun delivered to TIGR ran so hot that it shut down.
"The support relationship has really played a tremendous role," said Sapiro.
The outdated server system was also an issue from a research standpoint. It was "not just an operational problem and a cost problem, but it was actually impeding out ability to publish," said Sapiro.
The old computing system was simply too slow, a definite hazard in the rapidly evolving field of genomic research.
"An average microbial genome assembly could run up to a week or more," he says.
With the new server system "we were able to shrink that time down even on the larger genomes to half a day or a day," said Sapiro. "We have not had an assembly that would run anywhere near what it used to take."