Having experienced success with virtualizing servers, Aron Morgulis figured he'd try the same strategy with st...
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
Morgulis is director of information technology for the New York City-based American Committee for the Weizmann Institute of Science (ACWIS). Its primary function is recruiting philanthropic donations for Israel's Weizmann Institute of Science, a center for interdisciplinary research.
Managing data storage, on the other hand, is not its specialty. The 100-employee nonprofit needed a way to control storage costs, since adding new servers proved to be a law of diminishing returns.
Exploding cost of DASVirtualizing servers attacked one issue of hardware costs, yet another persisted: an ongoing need to add disk array after disk array to servers to keep pace with soaring data storage. Ideally, this meant attaching arrays with matching physical volumes of storage. More commonly, ACWIS found itself forced to attach whichever drives might be available -- even if it that meant coupling larger volume drives to smaller arrays, thus forfeiting tens of gigabytes' worth of storage capacity.
The only other alternative involved shutting down the entire network until new hardware could be purchased, installed and tested -- an unpalatable choice at best.
"We have a lot of different servers and different RAID subsystems for each one of them. If we wanted to expand the RAID sets to any significant degree, we had to have downtime," Morgulis said.
To solve these problems, Morgulis once again chose the virtualization route. For a small shop, ACWIS uses a variety of servers, operating systems and storage devices, including a 1 Gbit Fibre Channel (FC) SAN running EMC Clariion's FC 4700 arrays.
Except for certain utility applications at standalone facilities, the EMC-based SAN supports all major applications used by the organization's 12 offices. That includes an Oracle database plus applications powering Novell NetWare, SQL Server, and Microsoft Exchange domain controllers.
Morgulis went searching for virtualization technologies that would automatically ration storage as needed across the SAN while also supporting the various server environments. And Morgulis wanted something else: virtualized storage that didn't take a symmetrical path to the data. In other words, he wanted to avoid the prospect that each domain manager could become a single point of failure, and hamstring his entire network.
His five person IT team evaluated several storage virtualization products, ultimately whittling the field to two vendors. The first he won't identify, but said: "You know the name," probably FalconStor Software Inc. of Melville, N.Y. The second vendor was ACWIS's eventual choice: Irvine, Calif.-based StoreAge Networking Technologies..
ACWIS implemented the SAN Volume Suite, a combined product of StoreAge and Troika Networks Inc. of Westlake Village, Calif. It consists of StoreAge's virtualization software running on Troika's proprietary network storage platform. The selling point for Morgulis is StoreAge's use of "out-of-band" or asymmetric virtualization, a configuration that removes routing devices and virtualization devices from the data storage path.
Instead, a thin client gets loaded onto each server to virtualize storage. The StoreAge appliance manages the network from outside the direct data path. The innovation furnishes ACWIS with an adaptable storage infrastructure, according to Morgulis.
"As far as organizations like ours, this is avant-garde," he said.
Perhaps not for long, though. Analysts say the out-of-band design popularized by StoreAge likely will gain popularity since it avoids single points of failure, long an Achilles' heel for virtualized storage networks. Other vendors could begin offering similarly featured products. "Asymmetric virtualization is the key to creating single-box behavior across multiple arrays in a SAN," concludes a 2003 research note from Stamford, Conn.-based Gartner Inc.
Counting dollars -- and ROIIntegrating the SAN combination product with VMWare enables ACWIS to manage both servers and storage virtually, Morgulis said. As a result, the organization is free to concentrate on more important business matters, especially fundraising.
""We are getting more bang for our buck on storage because we've reduced our worry factor. We no longer have to fool with the hardware," he said. "Now if we want to add another system, all we have to do is put another host bus adapter into a piece of hardware. It doesn't have to be among the servers."
Morgulis declined to reveal his IT budget. No doubt he was shopping on a shoestring though. Getting the go-ahead to make necessary storage investments was no slam-dunk. Especially tough was convincing management to spend precious IT dollars on an expensive FC storage network.
"Let's put it this way: The first cost (estimates) I gave my boss he threw out," Morgulis said.
Back at the drawing board, Morgulis hit on a plan to buy used EMC FC 4700 arrays "at reasonable prices" on the secondary market. That more than made up for the licensing costs associated with purchasing the needed software from StoreAge, he said.
Drawback: No redundancyAsymmetry aside, Morgulis concedes his small SAN environment is not completely free of risk. Although removing routers and devices directly from the data path cuts down on the potential for network failure, it also eliminates redundancy. Morgulis said his enterprise simply isn't large enough to financially justify building multiple data paths. So the potential still exists for single points of failure, although Morgulis terms the risk "minimal."
For one thing, EMC provides built-in redundancy inside its storage boxes. To add capacity, ACWIS only needs to get an expansion for the box, stick in some disks and new virtual partitions would be created quickly.
Also, Morgulis said Troika has "prepositioned" backup storage devices at various locations within the New York City metropolitan area, meaning ACWIS could get a replacement for failed devices within hours.
"There is a certain amount of risk the way (this system) is made, but our business is not gong to stop if the computer stops working," Morgulis said.