Boot Storms Be Gone with Avere's NAS Optimization Solutions
I like keeping up on users, so when this case study from Avere Systems popped up, I was intrigued because it described a straight-forward solution to a vexing problem that more than one VDI system has encountered.
It all starts with the Belchertown School district in Massachusetts, which thought it was making all the right moves. Its IT team put together a system based on five Cisco UCS systems running VMware View connected to 10 data stores hosted on a NetApp FAS2020, and supporting one-terabyte volumes, with a 20 percent snapshot reserve. Most VMDK files are held to 40GB and are typically linked clones from one of a few golden master images.
This configuration supports PowerSchool, a virtualized application utilized by teachers and students alike to access class materials, log attendance and store and back up their work and grades. All in all, a locked-and-loaded system--until the boot storm ensued when some 250 concurrent users logged on at the same time.
It typically took 15 minutes for the storm to subside. During that time, students were unable to access their class documents, teachers got errors when they tried to save their work, and the school district's small, but valiant, IT team couldn't keep up .
They evaluated bigger, more updated NetApp filers, but the costs were prohibitive, and the time required to implement them was unacceptable. The remedy was to be found at a VMware user group meeting in Maine, where they discovered Avere's NAS Optimization solutions.
It was a Eureka moment for Scott Karen, the school district's director of Technology. As he puts it, "Avere was able to get us an FXT Series node for evaluation almost immediately. We tested it, and found it eliminated the boot storm and turned 15 minutes of logging in per class into about three minutes with no more write errors or corrupted files. Overall latency is no longer an issue. And it didn't require me to rethink and re-engineer my storage network."
Karen moved quickly, and within three weeks of choosing Avere, the FXT 2500 two-node cluster was up and running. The nodes are configured in read/write mode during the school day, and at 6:00 P.M., the cluster automatically switches modes to read-only in an effort to support the district's existing backup strategy with Veeam. Twelve hours later, the cluster switches back to read/write mode for maximum performance during the school day.
The FXT 2500 clusters are reliable, they don't need tweeking, and Karen hasn't had to touch them since they were implemented. He refers to them as "magic boxes," adding "Because of Avere's tiered storage architecture, all future disk purchases can be lower cost SATA drivers rather than higher cost SAS, so Avere has not only solved our performance problems, but will be saving us money going forward."
It is, as they say, all good.
Posted by Bruce Hoard on 01/04/2012 at 12:48 PM