A Look at the Future of Internet Services for Student Residence Halls
by Chip German, VP for Information Resources/CIO
[This information was updated April 18, 2008]
Observers of higher education culture have said that the only thing that unites all campus constituencies is a common complaint about parking. For students in the residence halls of campuses across America, an additional common complaint might focus on Internet services.
At a November 2007 meeting of public higher education IT leaders in Virginia, every campus representative present except two reported extreme difficulties in providing sufficient Internet bandwidth to student residence halls to satisfy the demands of the residents. The exceptions among institutions represented were Virginia Tech and George Mason Univerity. In the case of Virginia Tech, its identification as a leader in national higher education high-performance network design is a clear university strategic (and investment) priority. George Mason benefits from being located in a place where high bandwidth access to the Internet is as much as eight to ten times less expensive than other areas of the state. The rest of us are battling capacity shortages.
Most college networks are designed for first-rate access to campus-based resources, but often--in order to preserve bandwidth for the most important institutional mission components (teaching, research and public service)--compromises are made that limit access to off-campus based resources on the Internet. Both the Internet in general and our connection to it are engineered (and paid for) based on an old concept--that most Internet traffic is "bursty" in nature, rather than imposing long-duration loads. "Bursty" traffic is traffic that is uneven in flow, with occasional peaks of volume. But much of the nature of student use these days can be in the form of long-duration, high-volume loads (sometimes associated with the activities of an individual student, but more often due to aggregated student-use patterns), because of large file downloads and multimedia streaming content.
Current conditions at UMW
The University's connection to the Internet, shared between both the Fredericksburg and Stafford campuses, is a DS3 44.736 Megabit-per-second connection (via Asynchronous Transfer Mode--ATM) from which some 10 Mbps are reserved for academic and administrative buildings and the remainder is available for use by the University's residence halls. Our connectivity for this purpose is provided by NetworkVirginia at a cost of more than $53,000 per year, paid from a combination of state funding and tuition revenues.
Student demand in the residence halls regularly exceeds our connection's capacity. The recent trend, to no one's surprise, is that the favorite Internet-based resources for students are increasingly those that consume the greatest chunks of "bandwidth," such as YouTube, other video-distribution sites, and graphically intense "virtual-world" sites. The growth in bandwidth consumption in the residence halls has been explosive. Because bandwidth is expensive, the University applies significant effort in bandwidth management with this goal: manage the supply sufficiently to provide good service through short-duration bursts of traffic volume that would exceed capacity if not managed. Unfortunately, as I noted above, major spikes in traffic are no longer of short-duration, so our management strategy has in recent months not met either our students' expectation for service (especially at the times of day when their interest in access to the Internet is highest) or our own.
This is a complex challenge, so some additional explanations of how we have addressed it and how we will address it in the future are in order.
We are charged with the task of taking reasonable steps that our users are using our resources responsibly, and part of our management strategy is to use technical means to give priority for uses that are of the highest importance to the University. So the two factors currently determining the nature of student experience in residence halls here are supply (how much bandwidth we get from our suppliers) and management of distribution (how we make that bandwidth available to our populations). Let's start with the latter.
The Bandwidth Management Strategies
Our networking staff constantly analyze the patterns of Internet-related traffic at UMW. It will come as a surprise to no one that we often completely fill our connection's capacity in the evenings. One interesting feature of our analysis over time is that we have discovered incidents in which one user was personally responsible for consumption of 20 percent of the entire allotment of bandwidth available to the residence halls for extended periods of time, and other incidents more recently where individual users have consumed more than 29 percent for shorter durations, even with our bandwidth management strategies in place. Each of those cases required adjustment of our technical environment to prevent them from happening again. And historically we've seen evidence that much of the consumption of bandwidth has been related to misconfiguration of user software that allowed the user machine to appear as a file server to the Internet when using peer-to-peer applications--that means that much of the University's bandwidth was going to external users who were downloading large media files from student computers. This misconfiguration of computers is especially risky for students if there are questions about the copyright status of files that are visible through those means--it is a prime mechanism used by copyright-policing entities such as RIAA identify users to whom to send violation notices. Again, technical adjustments in our environment have helped limit repetition of the service effects of such incidents.
At UMW, we use "intelligent traffic shaping" to ensure that the uses of the Internet that the institution has identified as important can continue during peak Internet-related-traffic periods. By long-standing policy, rankings of the importance of specific uses fall into these broad categories:
- Highest and Primary: To support the education, research, and administrative purposes of the University of Mary Washington.
- Medium and Secondary: To support other uses indirectly related to the University of Mary Washington's purposes with education or research benefits, including personal communications.
- Lowest and Least Important: Recreation and entertainment.
It is fair to say that peer-to-peer networking (P2P) has been associated much more with entertainment and other non-academic activity than academic activity, and it resides in the "lowest and least important" category, despite intense interest from students.
We also limit the use of some software "ports" at the "transport" layer of the network when they have been identified as known targets for attacks on networks. For obvious reasons, we will not offer detailed public information on these security-related configurations.
Finally, we don't allow some services at the "application layer" of the network (often only in specific circumstances) when they represent known security vulnerabilities. For example, we have disabled the capacity of many so-called POP e-mail services to interact with UMW servers, because the POP services usually transmit passwords over the network in accessible, unencrypted form. University users may still use unencrypted POP with external servers if they wish to take that risk , but we only enable secure POP (encrypted) for use with University servers. And we have disabled SMTP (a common Internet standard for e-mail transmission) from computers in the student residence halls because of the frequency with which that protocol is used in generating spam addressed throughout the Internet from virus-infected or otherwise compromised computers.
Effects of Our Bandwidth Management practices
Despite the honorable goals of effective management of a limited resource on the University's behalf, our practices to manage bandwidth have the obvious consequence of irritating student users who simply want the University to provide the same level of service they had (and have) at home. A few years ago, that was an easy target to exceed for University technology organizations, but the intervening years have brought both increasing intensity of engagement in on-line interactions on the part of the students and much bigger costs in terms of bandwidth on the part of the traffic they generate. And in a world where the boundaries of the classroom and of academic life are increasingly being erased by greater use of multimedia and social networking tools in teaching and learning contexts, it is no longer accurate to say that these high-consumption uses of the Internet aren't part of academic life. The challenge, then, is to figure out how to build and finance an infrastructure to support them.
The immediate problem: Supply is too little at present
Even the most effective bandwidth management can't solve performance problems if the supply is simply too small. Because we have exhausted all of the available bandwidth management strategies and demand continues to grow daily, we have been working for the last several months to identify the most cost-effective options and to acquire funding for increasing our overall connection capacity to the Internet.
Although superficially simple, this solution has two complicating factors – it costs money in a budget year when cuts have been imposed from Richmond (with more on the way), and to accomplish the solution in the most cost-effective manner we need Richmond’s permission to directly negotiate with Internet bandwidth providers, rather than simply use the existing state contracts. The technical choices in buying additional bandwidth are also complex, requiring good analysis (which is now complete) to ensure the choices we make immediately are wise for the University over the long term. We also are happy to say that state officials have now granted the University authority to directly acquire the connections necessary to increase our Internet bandwidth, and work has begun to formally select a vendor to provide those connections at reasonable cost (the deadline for bids was April 18). Although we are hoping to select that vendor within days, all such providers need time to "build" their parts of the technical environment that we need -- usually between 60 and 90 days. That means that the improvements for which we've been working are unlikely to be in place before the end of classes this semester.
Longer term: The Need to Rethink our Approach
This topic has been a subject of regular discussion on the part of IT leaders at higher education institutions for some time, and about two years ago we at UMW began to look into options for dealing with this challenge. The most obvious starting point is a simple admission: dorm life is life. We have to manage the rest of the network for academic and administrative purposes, but the residence hall network will never again be manageable by those rules after students have experienced high-speed connectivity in their homes. If we are willing to admit that, we must admit another point: managing networks for broader life purposes, including entertainment, is not an area of core competence for higher education IT organizations. Years ago, we pioneered the networking of residential environments. Now, commercial entities have overtaken us--they offer wider ranging, less constrained and faster evolving network-based choices than higher ed IT organizations will ever be able to do, at least at an institution of our size.
Possibility of Turning Residence Hall Networks over to a Commercial Provider
With those two admissions more than a year ago, we started looking at the experience of other colleges, such as Ithaca College in New York, in outsourcing the provision of residence hall networking to commercial providers. The stories that we've heard have been positive ones, with much greater satisfaction on the part of students and greater capacity on the part of the institutional IT organizations to turn their focus back to the academic and administrative environments that must be managed by technical staff deeply connected to the institution's mission.
Last year , we studied the technical and financial feasibility of outsourcing residence hall networks. The University has now released a “Request for Proposal” for commercial providers to indicate how they would provide services to our residential-student environment. The RFP is available for viewing at
http://www.umw.edu/purchasing/vendors/bids_proposals/documents/RFP08-25.pdf
We will include student representation (and Division of Student Affairs representation) on the committee that reviews those proposals later in the spring. We are still hoping for an implementation in time for next fall, although the extended procurement processes needed for such a significant change may push that date further into the academic year.
FOOTNOTE: Some confusion exists about the role of Cisco's "Clean Access" product in the student and wireless environments. Clean Access is a tool that assists with authentication of users to the network while ensuring that their computers have received current operating system patches, and current virus definitions for their anti-virus applications. It performs no other functions in our environment and certainly has no role in traffic shaping.