The Storage Challenges of Desktop Virtualization

Martin Male thought he had found the answer to his latest IT problem in a Facebook ad, of all places. The manager of IT services at Yellowknife Education District No. 1 had introduced a pilot program at one of his schools to test out virtual desktop computing. But with the implementation of that, suddenly, storage had become prime infrastructure real estate. He was looking for a way to monitor and manage better the use of space on the district's storage area network (SAN).

But nine months into use of the new product advertised in that Facebook ad, the vendor he was dealing with made a right turn and dropped support in order to focus its business on cloud storage management. Now he's left searching for a replacement and still struggling with storage.

Virtual Computing and Exploding Storage
With 1,800 students and about 250 to 260 staff members, Yellowknife is the largest district in the Northwest Territories, a federal territory north of the Canadian province of Alberta. Its K-12 schools provide instruction in some combination of French, English, and Dogrib, the latter a First Nations language.

The activities that go on there are pretty much the same ones going on at schools in most parts of North America. The schools have computer labs; the computers run Microsoft Office applications; and the data center's physical servers have been virtualized for efficiency's sake.

About two years ago Male turned his attention to a virtual desktop scenario. The advantages are many, he pointed out. The virtual environment allows users to access their school applications and documents wherever they have an Internet connection. Also, as long the machine can turn on and access the Internet, it can be used for a virtual desktop. "We used to evergreen our machines in our labs every four years," he said. "Now, rather than replacing that physical lab every four years, we can keep those computers in the labs, put that investment into mobile carts, and [continue] increasing the number of computers in the schools."

But when Male implemented a Wyse XenDesktop virtual desktop environment as a proof of concept at one of his K-8 schools, he found himself grappling to dedicate sufficient storage capacity to that one school.

"In the one location that we have the virtual desktops, we run a SAN that has 6 TB of disk space," he explained. "That 6 TB SAN has 11 servers running on it, plus it has 80 student desktops, and 20 staff desktops. The other schools that are roughly the same size all run with about 500 GB."

The reasons for that added storage demand is twofold, noted Male. First, there's the storage required to contain virtual desktop profiles for each student. Second, as students and staff became accustomed to working virtually, they rely less on maintaining media on USB thumb drives and more on stashing their data on the network SAN.

"The kids are used to bringing the USB sticks to school to save work on," he said. "I've seen hundreds of those get zapped out by static electricity. We're bad in the winter time with static--we're considered the 'north desert' in the winter time because we have less than 10 percent humidity, so it's easy for static to zap things out."

With the virtual desktop setup, students don't have to use a USB drive. "They can go home, log into virtual desktop over the Internet and have access to that same data and be able to use the same programs they're using at school," Male said.

The result, however, was a SAN crammed full of content nobody was necessarily accessing anymore or that was misfiled. For example, each school has a server dedicated specifically to storing videos and photos generated in class projects. "I don't want photos taking up expensive space on my file server," Male insisted. "The photo servers use an older and cheaper server for that purpose."

At the same time the virtualization effort was going on, the district forced through policies requiring staff to use its file servers to save their data for backup purposes rather than relying on the local hard drives in their computers. "When we did that, there was a lot more data out there than we originally thought," added Male. "We had schools using 50 GB going up to 250 GB. We had two locations that were running out of space very quickly. We needed a way to be able to look at those locations and get some sort of information about what it was that was taking up the space and how old that data was."

While searching for a solution to the district's problem, Male happened to notice an ad on a Facebook page for a service named Aprigo Ninja. Ninja would allow him to know what kinds of files had been added to the storage recently or hadn't been accessed over a specified period. He contacted the company and tried out a demonstration. "They gave us exactly what we wanted," Male said. "We could run a weekly scan on file and print servers to tell how much space was being used, what kind of data was on them, and how old the data was."

Convinced of its value, the district signed a three-year license. Male set up a server specifically to act as a collection server for the service. That collection server interfaced with the Aprigo Web site, where the actual data analysis was performed. Then Ninja would produce a weekly analysis report on each server that Male would receive and review every Saturday. Those reports enabled him to identify content that hadn't been accessed in two or three years or that shouldn't have been stored where it was located. Using that analysis, he said, he could go back to the data owner and say, "This stuff is old. You haven't used it in three years. I'm going to burn it to a CD or a DVD and take it off my file server."

In the two locations where the district was running out of space quickly, Male was able to determine what was consuming capacity. "We were able to save ourselves probably 20 to 30 percent of the space off the servers by removing unwanted file formats and removing outdated files that hadn't been accessed in quite a while," he said.

Easy Storage Management Out, Virtualization Still In
Then in February 2011, nine months into his use of the service, Male received e-mail notification that Aprigo was changing its name to CloudLock and its business strategy to focus solely on cloud computing, where it was seeing massive growth. Its analysis services would be focused on helping organizations using cloud-based applications to manage their cloud-maintained data. The kind of service Yellowknife was using--to analyze content on-premises--would no longer be available starting in another month.

"I was a little ticked that they had done this," Male recalled. "We had paid for a three-year license on top of it. I asked for a refund of our entire three years. And they actually did send it to us." But, he added, "I'd still rather have the product."

Male has spent the last two months looking for a replacement--with no luck.

That doesn't mean, however, that his ambitions for running virtualization on desktops will be slowing down any time soon. Now in its second year of the pilot, the district has worked out its virtual desktop kinks and will, according to Male, to forge ahead when funds are available, possibly in time for the next school year.

One problem was relying on virtual desktop profiles to be fed from the district's central server farm. Sometimes the network link would stop or bandwidth would drop, which meant students and staff couldn't access their virtual desktops. To overcome that challenge, the IT organization set up a server farm at the school, consisting of three servers and a dedicated SAN.

That's the same model he'll follow in the other K-8 schools where he will implement virtual desktops.

"It's a huge expense. By the time we're finished doing our six schools, we'll have spent half a million dollars," he said. "However, we also 'evergreen' all of our lab computers every four years, and that costs $80,000. That won't need to be done."

Likewise, IT staff time will be optimized. For example, the computer lab running the virtual desktops recently added one program, which meant the image being fed to those profiles needed to be switched out for another one. Making that update in the old-fashioned way, machine by machine, would take about 20 hours, he estimated. Creating that updated image for the virtual desktops and pushing it out to all of the computers took a total of about 40 minutes. "If a student was still on," Male added, "they'd be working with the new image; when they logged off and back on, they'd have the new image."

The lone traditional high school in the district will follow a different model. There, many of the students and staff prefer to work on their own devices. Also, use of the computers tends to be more resource intensive, requiring heftier processing power, and Male isn't sold on the use of a virtual desktop in that scenario.

So users will be able to bring their own computers, tablets, and iPads and gain access to the Internet through a guest network. But when they want to access school resources, such as newer versions of software or applications they don't have on their own machines, they can pull up a Web browser and get into their virtual desktops. "While their device is still off of the network, they'd be able to access the inside of the network using that virtual desktop. That keeps our network safe and the kids get to use their devices at the same time."

On the storage front, Male said, he knows he'll be able to delete student files at the end of the school year, but there's still next year and the years after that to address. "We'll continue to look between now and the start of the next school year for a replacement [to Ninja]," he concluded. "Storage management is still a challenge. Storage is still expensive."

Featured