Michael Barmada
Apple SAN Solution Meets Massive Storage Needs
The departmental storage system can also grow without being taken offline. Thanks to Xsan, Barmada can add one or more Xserve RAID units — and thus more storage space — without interrupting the cluster while it works. Xsan simply incorporates the added storage and its volume grows. The cluster gets more space to hold queued jobs and users get a larger volume to store their data.
To link the SAN together, Barmada chose the Emulex 355 SAN Storage Switch. The 12-port Fibre Channel switch fits neatly in a standard 1U rack and supports up to 200 MB/s port speeds.
Setting up the Apple Xsan software was a cinch. “The traditional Mac ease of use has been ported to an application that has been traditionally not so easy to use,” says Barmada. “With other solutions, I was looking at weeks of training for my IT people. They didn’t have to go through any training for Xsan. They opened the program and started fiddling with a few things here and there and they were able to put the whole thing together.”
Vastly Improved Backup Strategy
The university has two mirrored Xsan storage pools, one at the colocation facility and another on campus. When the cluster system was first installed, Barmada had to find a way to copy several terabytes worth of files with a minimum amount of downtime. Mirroring the drives over the VPN was out of the question.
“With other solutions, I was looking at weeks of training for my IT people. They didn’t have to go through any training for Xsan. They opened the program and started fiddling with a few things here and there and they were able to put the whole thing together.”
“We took all of the data in our department and backed it up onto the older NFS servers there,” says Barmada. “Then we moved those to the off-site facility and mirrored them to the second SAN using Xsan. Now we have two SANs, one in our department and one off-site. We do nightly mirrors using a very careful application called Psync. It takes a long time to do a mirror, but if we’re careful about how we do it, it doesn’t generate a lot of network traffic and doesn’t bother the system too much. Now that we have everything replicated at both sites, there’s not a lot of changes that occur in one day — usually about 4,000 to 5,000. The backup process takes around six hours.”
If any one server goes down, the cluster won’t crash. The system can lose a hard drive or server without going offline — only the jobs that were running on the bad server or drive are interrupted. That’s a vast improvement over the old system. “If a server crashed, the whole system would error out,” says Barmada. “We’d have to start everything over again.” Now he can swap out one server or drive and get the system back to full capacity in no time.
More Data for Prevention and Cures
“For human genetics the ultimate goal is that everybody has their entire genome sequenced and filed when they’re born,” says Barmada. “So when we need to go do a research study, instead of collecting 100,000 or 500,000 markers, we’ll have three billion markers, we’ll have the entire genome sequenced. Then we can start to model what is actually going on in the genome.”
To keep pace with the enormous growth and complexity of data, biologists will need technology to stay ahead of the curve. “I see the need growing exponentially,” says Barmada. “The more I add, the more they use.” The current cluster is running at capacity and the available storage is being rapidly consumed. Barmada has already ordered another 2.8TB Xserve RAID to help scale the system with demand. In an effort to build resources for the future and anticipate future needs, the professor is teaming up with other departments in the university and other schools throughout the region. In exchange for a new Xserve G5 or a monthly fee, they’ll get processing time on the cluster for their own projects. This system of sharing resources will foster a free flow of scientific ideas and help further research in all of the university’s departments.
