University Boosts Network and Storage Performance

Managing IT infrastructure is an increasingly challenging task. What’s more, as organizations wade further into virtualized environments and attempt to optimize performance, the complexities grow significantly.

One entity that had recently confronted this issue is the University of California, Irvine, which serves more than 31,500 undergraduate and graduate students and has a staff of more than 2,600.

“We found ourselves dealing with an aging infrastructure for our storage, while the size of our VMware infrastructure was growing,” says John Ward, infrastructure architect for UC Irvine. “In many cases, performance was lagging, but we couldn’t pinpoint what was causing the problem.”

In fact, in some instances, systems were unbearably slow, data backups didn’t work right and latency levels for various tasks were completely unacceptable. The IT department, which oversees nearly 900 virtual servers, recognized that it had to transition away from older and slower hard drives to hybrid devices that incorporate flash storage and could operate within the virtualized environment faster and more effectively.

After surveying the vendor landscape and examining a number of directions and options (Ward formulated a vendor questionnaire with 92 questions) over a six-month span, the school turned to Tintri, which specializes in VM-aware storage (VAS) systems.

UC Irvine switched to the new environment in March 2015. The transition went smoothly, Ward says. “Because everything is in VMware, it was simple to set up two Tintri VMstore T880 storage devices without the computers going down,” he explains. “We just moved over to them on a live basis, and it wasn’t apparent to anyone using the systems.”

The university has since added two more storage devices. Ward says that as storage requirements expand, the university will continue to add units in the future.

Network Latency Down, Database Access Up

The results have been impressive. In the past, network latency was normally around 20 milliseconds or less. Today, it’s down to the sub-two millisecond range.

“We no longer have databases going offline and unavailable and batch operations completely interrupted,” Ward reports. “It’s no longer taking three or four hours to complete batch tasks. We’re typically down to around 30 minutes on these jobs.”

Likewise, various user systems are operating faster, and backup windows have shrunk. Ward says that the IT team has a great deal of flexibility in how it manages logical unit numbers (LUNs) and carves up storage blocks. “The technology delivers an entire abstraction layer that greatly simplifies things,” he notes.

In addition, the IT team has full visibility into network and storage performance through the Tintri Global Center, which serves as a single pane view of the storage devices. For instance, when a virtual server recently began underperforming and pulling 7 gigabytes of bandwidth on a 10 gigabyte connection, IT staff members were able to isolate the virtual machine immediately and fix the problem.

“In the past, we would have had no way to know where the problem was originating and how to isolate and fix it,” Ward explains. All of this has led to time savings for network administrators and an ability to use their services in more strategic ways.

“This approach has made an enormous difference in how we approach storage and how we deliver services,” he reports.