Oakland University (in Rochester, Mich.) has a strategic plan that emphasizes a strong academic mission and encourages the growth of the university. University Technology Services (UTS), the central IT department, has to plan carefully to support the university’s strategic goals, including plans to expand enrollment to 25,000 students (with corresponding growth in faculty and operations) by 2020. A growing population leads to increased demands for storage and server-based services, and virtualization is helping us meet those demands.
In the course of the university’s annual asset-planning review, an analysis of the number of servers, storage devices and racks led us to believe that floor and rack space were obstacles to growth. We didn’t think we could continue to fit servers and storage devices into one room.
When we began planning the data center, we estimated we would have ample floor and rack space for 10 years, given the current growth rate. Upon closer examination of related environmental support data, however (especially the data center’s electrical load and cooling systems), we found we were on a tighter deadline.
We determined that our environmental systems would support only a few more years of growth at the current rate. Further, cost estimates were high for capacity improvements in these areas. We wanted to find “green” alternatives that would reduce our power consumption, and thus the environmental footprint of our data center.
As we were planning, the director of systems engineering in UTS was seeking improvements to the department’s systems administration and operation. The director’s goals were similar to central IT’s ideals of high availability, reliability, speed of implementation, implementation consistency and scalability. Both our operations needed to be compatible with the existing infrastructure. A review of storage environment technologies led the group to consider virtualization.
Virtualization creates a system as a subset or partition of a larger system; it makes a single large system appear to function as several smaller systems or resources. Virtualization includes a processor, disk, memory and operating system. Instead of buying a new server for every new service or application, we could implement a virtual server on an existing platform or structure.
The virtualization environment supports templates that provide fast, standardized and secure server implementation. Once a template is built, we can attach the system to storage best suited to the system’s purpose, allowing for ease of expansion. All the drives, disk arrays, head nodes and Fibre Channels are redundant, providing for fail-over between heads for easy updates, with no impact on end users. The need for compatibility with our existing hardware led UTS to select VMware as our virtualization platform.
Training, setup and creation of initial templates took about a month. We are now engaged in P2V — taking our existing physical servers and moving them to virtual servers. VMotion supports an entire system migration that allows for updates without any end-user downtime. This requires checking with the application vendor to make sure its system is supported in a virtual environment. New systems are also checked at the point of purchase.
The result is that fewer servers are purchased, which improves our environmental footprint by reducing heat generation and electrical requirements in the data center. Installation of new servers is much faster, allowing the IT staff to be more agile and responsive to growth in university services. The virtual environment is more fault-tolerant, improving reliability.
The downside to virtualization has been minimal, but there are a few issues. A single power outage on a power strip can affect multiple virtualized systems. The technical architecture is more complex; support involves not only operating system administration but also virtualization administration. Planning and management can help minimize the risks from these, however.
Overall, Oakland University’s move to virtualization has improved our green footprint while supporting university growth. The topology supports innovation, agility and responsiveness, providing a much improved infrastructure.
Stephen Glowacki is director of systems engineering at Oakland.