At William & Mary, our IT infrastructure team needs to be nimble enough to support a leading-edge research university — and deliver the stability expected of a 325 year old institution. We’re not a large school, but we have a long history. We’re a public university located in Williamsburg, Virginia, and founded in 1693, making us the second-oldest institution of higher education in America. Our alumni range from three U.S. presidents to Jon Stewart.
The Linux team in the university’s central IT department is made up of 5 engineers. We run web servers, DNS, LDAP, the backend for our ERP system, components of the content management system, applications for administrative computing, some academic computing, plus a long list of niche applications and middleware. In a university environment with limited IT resources, legacy applications and infrastructure are expensive and time-consuming to keep going.
Some niche applications are tools built by developers in university departments outside of IT. Others are academic projects. We provide infrastructure for all of them, and sometimes demand can ramp up quickly. For instance, an experimental online course catalog was discovered by our students during a registration period. Many students decided they liked the experimental version better and told their friends. The unexpected demand at 7am sent developers and engineers scrambling.
More recently, IT was about to start on a major upgrade of our ERP system that would traditionally require at least 100 new virtual machines to be provisioned and maintained. The number of other applications was also set to double. This put a strain on our network and compute infrastructure. Even with a largely automated provisioning process, we didn’t have much time to spare.
We wanted to tackle both the day-to-day infrastructure management challenges and the scalability concerns. That’s what led us to look at Docker. After successfully running several production applications on Docker Engine – Community, we deployed the updated ERP application in containers on Docker Enterprise. We’re currently running on five bare metal Dell servers that support 47 active services and over 100 containers with room to grow.
Docker Enterprise also dramatically simplifies our application release cycles. Now most applications, including the ERP deployment, are being containerized. Individual departments can handle their own application upgrades, rollbacks and other changes without waiting for us to provision new infrastructure. We can also scale resources quickly, taking advantage of the public cloud as needed.
Just like our researchers have done for years, Docker has also enabled deeper collaboration with our counterparts at other universities. As we all work on completing the same major ERP upgrade we’re able to easily share and adopt enhancements much faster than with traditional architectures.
Today, Docker Enterprise is our application platform of choice. Going forward, it opens up all kinds of possibilities. We are already exploring public cloud for bursting compute resources and large-scale storage. In a year or two, we expect to operate 50 percent to 80 percent in the cloud.
Guest Blog | Phil Fenstermacher, Systems Engineer, William & Mary