|
Bioinformatics researchers generate a lot of data, especially those involved with some of the newer "-omics," says Dustin Machi, computational facility manager at the Virginia Bioinformatics Institute (VBI) at Virginia Tech. "Genomics has been out there for a long time and [researchers] might generate megabytes per day. But new fields like proteomics and metabolomics generate gigabytes and gigabytes per day," he says.
VBI researchers used to store those gigabyte-sized--and in extreme cases, terabyte-sized--data sets on IBM's Enterprise Storage Server (ESS) or "Shark" storage, but Machi and his colleagues are actively encouraging researchers to tap into VBI's new storage resources like iSCSI arrays from Rackable Systems. Thus far, VBI has deployed approximately 12TB of raw capacity across Rackable S3009 and S3016 arrays, which run iSCSI target software from Wasabi Systems. VBI plans to add many more terabytes of iSCSI storage arrays over the next couple of months.
It shouldn't be too hard to convince researchers to use the iSCSI disk. While researchers pay $4.50/GB per month to store data on the ESS, Machi can offer them iSCSI storage for approximately $3/GB per year, or a measly $.25/GB per month--one-eighteenth the cost. Furthermore, the iSCSI arrays are mirrored to a remote location and the Shark isn't.
The high cost of storing data on ESS had prompted some researchers to store data on local IDE drives they purchased themselves. But that was done at their own peril. "We'd tell them beforehand, 'We're not going to feel too sorry for you if you lose all your data,'" says Machi. Unfortunately for some researchers, Murphy's Law struck and data was lost.
Now, with iSCSI storage in place, VBI researchers should be able to get protected storage for less than the cost of IDE drives, predicts Machi.
From an administrative perspective, Machi has found that he can deploy iSCSI storage to a server much faster than with Fibre Channel (FC). Because VBI's servers run only Linux software initiators, there's no worrying about whether or not the server has a host bus adapter. There's no need to secure a route through the FC switch, and you don't need to reboot the server for it to recognize its new disk.
There are some downsides to iSCSI storage. As more and more hosts contend for iSCSI storage resources, the Gigabit Ethernet link serving up storage may start to become saturated. And Machi sometimes feels somewhat alone deploying iSCSI.
"The hardest part for us is that there aren't very many people doing it," he notes. And even though Rackable Systems and Wasabi have both been helpful, "it's not like with Linux--we're doing this without a lot of support from the Web."
|
