[ Thanks to LinuxNews.com Editor for this
link. ]
“Amid the sometimes-hazy paradigm shift from the centralized
supercomputer to the distributed cluster model, VA Linux has, on
several occasion, teamed up with computing superpower IBM to create
deluxe superclusters and provide a turbo-charged, lower-priced
alternative to the tried-and-true supercomputer. So far, the
U.S. research community has been the main beneficiary of this
emerging trend in the Open Source community–corporate megalith
teamwork. Facilities such as Los Lobos and Chiba City have garnered
much media attention in recent months and showcased the direct
influence the Linux community is having on the broader computing
landscape.”
“What has the company learned from these massive deployments?
Quite a lot, according to VA Linux Senior Cluster Developer John
Goebel. In a recent interview with LinuxNews.com, Goebel emphasized
cluster computing on shared machines instead of what he called the
“metacompute-like” legion, or the data grid. Here, Goebel explains
why the distributed cluster is getting so much spin, examines
problems he’s experienced while wrangling the critters, and
discusses the Linux management framework….”
“There are several reasons the trend has shifted. An early
research book on distributed clusters from CalTech had the
cartoon-ish title of `It Works.’ This was a book that came from
research they did on early parallelism. They where able to show
parallelism on commodity workstations could be used for real
research. The first hurdle to get through is `can commodity
distributed clusters be used to serious research?’ Pretty much, the
answer seems to be `yes.'”