Texas Christian University (TCU) recently dropped Network Appliance, Inc. (NetApp) as its storage vendor of choice...
By submitting your email address, you agree to receive emails regarding relevant topic offers from TechTarget and its partners. You can withdraw your consent at any time. Contact TechTarget at 275 Grove Street, Newton, MA.
for Microsoft (MS) Exchange failover, tapping instead SteelEye Technology Inc.'s LifeKeeper software.
The move has produced significant savings -- tens of thousands of dollars annually -- in addition to helping make the school's environment more efficient and reliable, according to Bryan Lucas, server administrator for Forth Worth, Texas-based TCU.
Lucas said that years ago, the trouble in setting up safeguards for the school's approximate 2 terabytes of email data had much to do with how the native tools for MS Exchange are limited to clustering within a single data center. Across physically separated clusters -- as is their case, with a second location on campus for disaster recovery purposes -- recovery took hours of time. There were also real concerns about support from Microsoft and its ability to upgrade from Exchange 2000 to 2003.
Lucas said storage area network (SAN) replication over IP seemed an appropriate answer to connect between the primary site and "The Bunker," the school's underground standby data center connected via a 10 Gbit link. But after years of hiccups using hardware (NetApp Filer 820) and software (SME v1.1 for Exchange) from NetApp, it had a difficult time legitimizing the money they were spending.
"We used a SAN product for software mirroring, spent a whole lot of money and time, but it didn't work right," Lucas said. "Not to say the product can't do it. But it didn't scale our environment in the way we needed it to. I also don't think it was up to speed with Exchange 2000 at the time. It wasn't efficient with snapshots. There was a limitation on how many you could take. There was a lot of wasted space with deployment. We were tired and couldn't get anywhere with it. I don't want to give them [NetApp] a bad rep … but I got the impression they didn't do a lot of Exchange [customization]."
NetApp spokesperson Michael Moeschler declined to comment.
"It didn't mirror at the OS level, it just mirrored the data," Lucas continued. "If something happened and we had to failover, I had to bring up an Exchange server, connect it into the SAN and hook in the data. It was a two hour process at best."
Lucas heard about SteelEye at an Exchange conference and invited the company to his offices. He said the difference is in the programming of SteelEye's Lifekeeper software, which communicates well with Exchange and isn't designed for a specific storage platform.
"It's five grand for a pair of HP [Hewlett-Packard Co.] ProLiant DL380 G4] servers," which Lucas said was very reasonable. "We were spending six figures doing the SAN route. The cost difference was big enough where we didn't even measure it. It's cheaper to do local shells and our own controllers. The idea of optimizing storage with SAN is attractive, but it's just not cost-effective. We're 95% HP ProLiant. For the most part, we always have spare hardware, we can put another 380 in the rack, move the drive, and we're done. Not all the benefits are the same but certainly not the cost benefit."
In the name of disaster recovery, Lucas was prepared to give up some of the benefits SAN can offer: high disk I/O performance allowing swift user probes of common folders and network snapshots that enable administrators to essentially back up in time and access previous database states. But Lucas said this sacrifice was assuaged because of strong hardware in ProLiant that didn't require a SAN boost, and also because he discovered a way around snapshotting via the customization of native Microsoft tools.
"I don't really need snapshots anymore," Lucas said. "It's kind of irrelevant."
Still, the disaster recovery bases aren't all covered. With a real-time mirror between two servers, if the primary gets corrupted, it gets passed to your secondary. Data corruption can be the result of bad hardware, a flaky controller, a power outage and remains a source of dread for many a techie.
"We want to get a third server and partition it out, just have it lag behind by 24 hours or something," Lucas said, citing additional worry to how the area is prone to flash floods. "But I could dream up all kinds of other things."
In a space that is fairly mature, price and cross-platform flexibility are differentiators that can be deal clinchers, especially for an organization with a scattered environment, according to Joyce Becknell, research director for Sageza Ltd.
"The days of saying 'we're red and you're blue' are gone in this space," Becknell said. "But SteelEye doesn't have a hardware agenda, and people can have any number of different things. With EMC Corp. and IBM there's a tendency to say, 'how about putting it on nice shiny iron?' "
Let us know what you think about the story; e-mail: Joe Spurr, News Writer
This article originally appeared on SearchDataCenter.com.
Dig Deeper on Microsoft Exchange Server Storage and Data Management