The institute's gene-sequencing software was named to Information Week's Greatest Software Ever Written list on Aug. 14 as number three out of 12 on the list.
Sun's ability to place its x86-instruction set servers, the Sun Fire V40Z, in a demanding, scientific environment is one sign of why it's been able to restart server sales and renew its fortunes. By designing workstations and servers based on AMD's 64-bit Opteron chip, Sun has departed from its invented-here, UltraSparc mentality and adopted what's been winning, according to marketplace economics.
At the end of 2004, the institute purchased three Sun Fire servers and ran them alongside its existing 15 Alpha servers. The original gene sequencing software had been ported from Alpha to Linux in 2000, paving the way for the changeover. When the institute found its ported software produced the same results on Sun Fire, only faster, it switched off the Alpha servers earlier this year and let the V40Zs take over.
Sequencing tasks that used to take a month or more on the Alpha servers now take "a few days or a few hours," Sapiro says. "That makes a huge difference to the institute—to get the data out faster."
The institute paid about $30,000 per server for the Sun Fires, compared with $100,000 per server for the Alphas in 1999, Sapiro says. He estimates that his cooling and electricity needs have decreased 70% with the changeover and space has opened up in his data center.
The institute had to buy 64-bit systems in 1999, before they were commonplace, because of the gene sequencing software's need for huge amounts of address space. Most 32-bit systems can generate up to four gigabytes of virtual memory but that wasn't enough, Sapiro says. Alpha was an early 64-bit system.
Bob Friday, Chief AI Officer for Juniper Networks, explains how the advanced technology is transforming operations.
Contact center leaders from 8x8, Awaken Intelligence, and 360insight discuss the importance of agent experience.
AI may force enterprises to rewire parts of their data centers so they are fully optimized to run such workloads. The question is do you use Ethernet or InfiniBand?
