Before we could deploy new e-learning technologies to students, we first needed to design and implement a scalable digital campus infrastructure which would increase the number of computer access points on campus available to students. Mekelle University lacks general access computer labs for students, and the existing course-specific labs are often hampered by problems that are typically associated with shared, anonymous, fat-client computing environments: configuration problems, viruses, hardware failures, out-of-date software installations, etc.

Thus, the digital campus infrastructure was designed with these criteria in mind:

Network-centric architecture – Instead of using traditional PC’s in the labs, we are using various thin-client technologies for the computing desktop, and the “intelligence” or aggregate compute capacity of the university is held within shared, robust, redundant servers in the MU data centers. This “ultra-thin” compute capacity, means that the actual desktop devices can be made up of low-cost, resilient devices that can be used for many years beyond the end-of-life cycle of a typical PC. Because there is no software configuration stored on the desktop device itself, the staffing requirements and IT skills required to manage the actual labs is eliminated. All the compute power, complexity, and IT management is centralized in the MU data centers, which are distributed across two campuses.

Identity Management – Since the desktop thin-clients that are used in the labs are essentially “dumb terminals”, students are given individual, password protected user accounts. Students can differentiate their desktop environment and applications, and store their documents and personalized content without creating a dependency or affinity with the desktop device itself. In other words, a student (or faculty member) can sit at any terminal in any lab and access their personalized desktop environment and content. This eliminates the need for flash drives, CDROM support, and other forms of “sneaker-net” file sharing.

Access to the lab terminals is done in two ways. First, all the terminals run a locked-down browser in kiosk mode, for anonymous user sessions. This is for those users who ONLY want access to an internet browser for a short period of time. Secondly, users that have a system account can access an authenticated personalized session using a username/password combination and a physical smartcard. The smartcard allows students to quickly access an existing, running session, from any smartcard based terminal. This way, students who wish to leave applications running on their desktop, but do not need to have an interactive session, can free up the desktop thin-client for other students. When the smartcard authenticated user returns to the lab, he or she can sit at any terminal, insert the smartcard, and recover the live, running existing session without closing or canceling any applications. The application stack available to users are primarily open-source applications. These programs all execute on shared servers and shared operating system instances (UNIX/Linux) in the data center.

The compute and IT management facilities that are required for a network-centric architectures with identity management are not without their complexities and challenges. Whereas in traditional fat-client lab environments, the availability and maintenance tasks of desktop PC’s is distributed (and often times falls into the hands of the end-user), with network-centric architectures, the lab thin-client functionality depends entirely on the network and server availability maintained by the university ICT staff. In this project, a large portion of the training program for ICT staff is focused on these issues.

Since the network-centric design, and identity management facilities allow us to use low-cost, low maintenance, low-powered devices, we will be able to scale this solution quickly and cost effectively throughout the rest of the university and beyond. In our opinion, it is better to spend the lion share of our project budget on staff training and capacity building with tutors and ICT staff, rather than on expensive, and often times unnecessary equipment. To that end, the digital campus pilot explores four different inexpensive thin-client technologies: Sun Ray ultra-thin clients (mostly used, some devices up to 10 years old), PXE bootable PC’s with a linux kernel for remote desktop access, Sun Ray access client software on recycled PC’s, and LTSP linux stations with recycled PC’s. The data center services are built on a horizontally scalable platform, meaning that older, heterogenous servers can be added to, not replaced, and share the load with newer servers as they arrive.

As the number of students and tutors accessing moodle and the digital campus infrastructure increase, additional management and monitoring services are being implemented: data backup services, hardware, software, and network availability monitoring, user provisioning tools, problem and incidence tracking, help desk facilities, etc.