The goal of the University of Kentucky (UK) CC-NIE project was to leverage high-speed programmable networks to create a campus network infrastructure that enables scientific research not previously possible using conventional network designs.  Because the cloud and remotely accessed national research facilities have become critical to many scientific research problems, the network also needed to support high-speed access to regional networks and Internet 2.

The conventional approach taken by most institutions is to create a high-speed science DMZ network separate from the campus network infrastructure, reserved exclusively for scientific research, and populated with machines and services focused on research.  The UK CC-NIE project explored an alternative approach that leverages programmable software define network (SDN) technology to create a single dual-purpose campus network infrastructure capable of carrying standard campus traffic and high-speed science traffic simultaneously.  Nodes in the network do not need to be identified as research nodes or standard nodes before being attached to the appropriate network.  Instead, individual (science) applications running on any node in the network can dynamically select and obtain access to high-speed paths across the campus network when needed.  Standard applications not requiring high-speed network service are sent over conventional paths through the network.  Although there is a single campus network infrastructure, the SDN-enabled network infrastructure together with SDN policies provide virtual separation of traffic across a single network.

Our motivation for this design was three-fold. First we wanted to enable high-speed research network access to researchers with limited network knowledge and resources and allow them research network access from any access layer port in their research facilities. Second, we wanted to enable more flexible and advanced programming capabilities for those researchers with more complex needs so that more complex rules can route network traffic based on the specific application not the physical access layer port. Third, we wanted to extend SDN capabilities more broadly across campus leveraging existing infrastructure wherever possible. To do this cost-effectively required the merging of research and non-research network architecture. This helped us take advantage of existing infrastructure and multiple funding streams to sustain a mixed use network of scientists and users.

To create a single network serving both conventional users and domain scientists, we upgraded the existing campus network infrastructure in-place, replacing existing routers and switches with SDN-enable routers and switches to create an SDN network that reaches the desktop of all users in the affected buildings.  In particular, network routers and switches in the Computer Science building, the UK Laboratory for Advanced Networking, the Chemistry/Physics building, the Computational Sciences (supercomputer) facility, and parts of the College of Arts and Sciences building were replaced with high-speed SDN-enabled equipment, thereby reaching a wide range of domain scientists at UK and making research resources (e.g., the UK supercomputer and storage facilities) accessible via high-speed channels.  In addition, the core routing and switching infrastructure connecting UK to the KyRON regional network and Internet 2 were upgraded to improve performance to national research facilities.

To manage the new SDN-enabled network infrastructure, we developed software that was able to dynamically insert SDN policies and rules into the SDN controllers that defined how t...

Please report errors in award information by writing to: awardsearch@nsf.gov.