Storage Virtualization Helps Slumberland Sleep Better

While there are a lot of jokes that can be made about a company that focuses on selling mattresses and bedding, Slumberland -- with 118 stores in 10 states -- clearly isn't doing any snoozing. Based in Little Canada, Minn., the company bills itself as a specialty retail furniture seller and one of the top 10 bedding retailers in the U.S. With its stores spread across the upper Midwest, Slumberland has to deal with a lot of data, and it has to handle it quickly.

"What we've done is migrate the vast amount of our storage from proprietary dedicated SANs and DAS to two Compellent SANs," says Seth Mitchell, the Infrastructure Team manager for Slumberland. "One is a larger primary with a lot of redundancy. A smaller secondary SAN is for replication and the secondary corporate office."

The company needed to move to a new virtualized storage model for a variety of reasons, Mitchell says. Those reasons included performance as well as simplicity of management and storage efficiency. "The number one reason was that we were not able to manage the proliferation of storage before," he says. "We run a very lean operation. We have about 75 servers and about 50 TB of storage."

The company's existing storage systems and the lean operation were causing problems. "We had a legacy EMC Clarion and some HP SAN components," he says. "We had a lot of DAS, some external and some internal. We had a lot of little silos and were running from one to the next dealing with the problems of dedicated systems. We were having problems allocating storage, and making adjustments. We would find out we were wrong and have to expand the storage and that wasn't always possible."

Of course, there was more to the move than just realizing that it was needed. First, Mitchell had to show that his plan would work, and that it would be cost effective. This meant dealing with the storage issues in the content of other business IT needs. "We needed to do a core retail business upgrade and we needed to do an expansion of our Exchange deployment. We actually ended up using the Exchange as a proof of concept," he says. "We actually coordinated with the server migration. We brought the new servers up on the Compellent SAN and migrated data from the EMC and HP SANs. We proved it would work with Exchange and showed the extra features to get the go ahead."The proof included a stress test, and even a disk failure. "The biggest thing was the performance and reliability," he says. "We were able to demonstrate some features like proactive notifications. We ran stress tests, and burned out a disk and we were able to see the whole process happen. We were still running, we were able to get support. In fact they [Compellent] told us about it before we knew we had a problem."

The initial deployment was taken one step at a time. "When we first put it in we had a dual-controller, dual-switch system," he says. "We had some refurbished 1-Gig switches and an enclosure of 73-GB disks for about a terabyte. Then we added two enclosures with two Fibre Channel switches and 73-GB disks for total of 3 terabytes. Now we have a mix of Fibre Channel and iSCSI."

But as is often the case with storage, that was just a start. "On our primary system we have seven enclosures. Four of them [have] 16 300-GB drives, for about 20 TB of storage, with 4 Gbit/s Fibre Channel," Mitchell says. "We have three enclosures with 750-GB SATA drives for about 36 TB of storage. We're up to 56 terabytes on that system. The secondary system is 16 500-GB Fiber ATA drives for 8 TB. Overall we have 64 TB."

Now that he has everything in place, Mitchell is delighted with the results of his virtualized storage. "It's been great," he said. "There aren't a lot of things to say. Storage is a non-issue to us now. We have a lot of flexibility."

Using his combination of dual Compellent Series 30 controllers and his Compellent Series 20 controller, he's able to accomplish tasks there were simply impossible in the past. "We have the ability to dynamically expand volumes," Mitchell says. "We can copy Oracle server volumes to a new server and upgrade it from there. It makes doing things like physical to virtual conversions easy. We boot from the SAN so reliability has gone up. When we do have a problem, we simply un-map a server and map a different server and we're back in action.""The ability to expand capacity is something that you really can't overstate," he adds. "We've gone up in the number of disks and number of spindles. We run into situations in reporting where we've had to add the capacity online without even having to reboot the server. We can remount the drives at speeds they couldn't imagine."

Virtualization came at a lower cost than Mitchell originally expected. Partly this was due to saving money on infrastructure, and partly because of the cost and efficiency of the Compellent virtualization solution. "We probably spent $175,000 total for the storage upgrades," he says. "We really haven't figured the ROI improvement. You'd have to put a price on the problems we had. We couldn't handle larger mailboxes; people were waiting for the server to respond. There are things that changed so dramatically that we would not be able to quantify them. Some processes used to take six hours now take six minutes. Its hard to quantify."

While Mitchell is very pleased with the success of his storage system, he's not done. Now that his primary storage environment is stable and running, he's ready to start on the secondary site. "The next step is expanding our secondary system to get closer to parity with our primary system in case we have something catastrophic at our primary site," he says. "It happens with us one step at a time."