Because “faster” means “better” in an efficiency-obsessed culture, businesses have embraced virtualization technology, which makes it possible to quickly and cheaply connect a lot of people and information over a broad array of devices. Following this trend, Cornell has set up its own infrastructure virtualization program to reduce its spending, open physical space and create a stronger IT infrastructure, according to University officials.
Through virtualization, businesses can package applications up as “virtual machines” and ship them off to a cloud computing data center at sharply discounted rates. The technology, which allows businesses to move their email and calendar systems from a local server to an online “cloud,” is leading groups like the University to ask themselves why they should pay more to possess their own equipment, according to Prof. Ken Birman, computer science.
“When we say cloud computing, we really have this bigger picture in mind,” Birman said. “The University move to Gmail is … [one] example of how Cornell has embraced the cloud for economic reasons.”
According to Birman, cloud computing has emerged from a combination of widespread Internet access and a huge wave of new network-enabled applications.
According to Andrea Beesing M.A. ’77, assistant director of unified communications and collaboration of IT@Cornell, infrastructure virtualization seeks to reduce Cornell’s overall IT costs and improve its service reliability by consolidating and virtualizing 80 percent of the distributed physical server infrastructure.
“The Infrastructure Virtualization Initiative was designed to save the University money by consolidating servers into the central data centers,” she said.
Last month, for example, as a part of its virtualization initiative, the University moved many of its tech services, such as the email and calendar system for about 24,000 faculty and staff, from a Cornell-provided local server, to an online-based cloud system, Beesing said.
“The migration of [Microsoft] Exchange [email] accounts to the cloud is part of the larger strategy to deliver utility IT services more cost-effectively in order to focus more IT resources on academic technologies,” Beesing said.
According to IT@Cornell’s website, through an increase in the number of servers a single staff member can manage and reductions in power and cooling costs, the University will save $2,845 per year for each physical server that is virtualized. Currently, the CIT-managed data center staff oversees 2,000 servers, according to Beesing.
Some units which have begun infrastructure virtualization projects include the College of Architecture, Art and Planning, Alumni Affairs and Development, the College of Agriculture and Life Sciences, the College of Human Ecology and Student and Academic Services, according to the University’s website.
Birman said he wrote to Provost Kent Fuchs several years ago urging the University to explore the benefits of virtualizing its systems.
“With virtualization, you can take a whole computer and package it into an image that will run on some other computer in an identical way,” he said. “We end up with one busy server hosting perhaps 10 or 20 desktop computer images or server images, and all those machines can be replaced with much cheaper, more energy efficient display-only devices. The benefits are really huge.”
The project was mentioned again last September, when faculty and administrators gathered for an all-day conference to give their input on the University’s strategic plan for IT@Cornell. During the conference, Ted Dodds, the University’s chief information officer and vice president for information technologies, emphasized the need for cost-saving initiatives such as virtualization.
“[The University has tried] to reduce the cost of those utility services that we all consume and need, but [so far, it] doesn’t actually differentiate the University from its peers,” Dodds said on Sept. 11. “But then [to] take the savings and the optimization we get from those utilities and reallocate to things –– that will make a difference.”
According to Birman, cloud computing, nonetheless, has its downsides, such as inconsistency in service and ineffective security.
“My main criticism of today’s cloud model is that the cloud has not been ideal for hosting sensitive data or applications, like medical records, systems to transform the power grid into a smart grid [and] self-driving cars,” he said.
Last month, for example, the Cornell housing portal crashed during the housing lottery because of the server’s limits –– a problem larger cloud servers could face as well, according to Beesing.
Original Author: Jonathan Swartz