Patrick Crowley, PhD, associate professor of computer science and engineering, has received a two-year, $570,000 grant from the National Science Foundation for research titled "MRI: Development of a Testbed for Overlay Network Research and Education."
The grant will allow the Open Network Lab (ONL) to add 'Virtual Hosts' via platform virtualization, add support for software-only virtual routers based on commodity servers, and enable the use of virtual appliances in ONL experiments.
The purpose of this proposal is to develop a controllable, high-performance experimental testbed for overlay network research and education. It builds upon and leverages an existing Internet-accessible community resource, the Open Network Lab (ONL, www.onl.wustl.edu
), adding high performance components that can be configured and extended to support a wide range of controlled and repeatable experimental computer networking projects. Over the past year and a half, ONL has experienced both a dramatic rise in usage and a significant shift in the nature of the experiments conducted by its users. In particular, ONL is being used to prototype and evaluate overlay networks, peer-to-peer systems, and novel future Internet architectures, and we have observed that these experiments typically seek to use as many resources—routers, switches and servers—as possible; it is routine for a single experiment to consume 60% of the available resources for the duration of its experimental reservation. ONL has been effective in supporting these workloads, and this fact has driven the demand for its resources amongst a growing set of users. Through the integration of additional switches and servers and the addition of platform-level server virtualization, this proposal seeks to expand the available physical resources in ONL by a factor of two and, via server virtualization, to expand the effective virtual resources by a factor of approximately eight. Beyond increasing capacity, this development will remove long-standing limitations and enable important new uses and research activities in ONL. To date, ONL user accounts have not been granted root access except for a few individual exceptions. The need to maintain a consistent ONL-compatible Linux image across all hosts has been the roadblock. By defining a VM-based ONL-compatible system image, the mandatory characteristics of the server platform can be separated from the details of the guest operating systems chosen by users. This in turn grants users the ability to configure all aspects of the guest OS, including the installation of packages, choice of OS kernel and choice of kernel modules. This capability will enable users to evaluate and develop systems software, as well as VM appliances. The increased scale and diversity of the ONL testbed components will directly support the research and education activities of our large and growing user base.
This is an infrastructure development project, seeking to create an important community resource. Its Intellectual Merit lies primarily in the research and educational activities that will be enabled by the availability of the new capabilities being developed. By providing an experimental environment supporting experimentation in future Internet architectures, overlay networks, and peer-to-peer systems, the testbed enables researchers and students to pursue projects that are currently outside their reach, because they require an experimental infrastructure that they cannot afford. The testbed also makes researchers more productive, by providing the necessary tools and infrastructure to allow them to focus on the innovative elements of their research ideas, freeing them from many of the more mundane tasks that typically have to be done before the more interesting issues can be addressed.
The Broader Impact of the program will be felt in several ways. First, by enabling better experimental evaluation of networking research ideas, the testbed can help to contribute to their development and can facilitate their transition from the research context to practical application. Second, we expect the educational component of the testbed to have a significant impact on undergraduate and graduate education. Students who have meaningful laboratory experiences can make a more direct connection between networking concepts and actual network operation, allowing them to learn faster and become more productive. Third, we believe that this project can contribute to attracting more students to careers in science and engineering. The growing impact of globalization on the world economy makes it essential for the United States to continue to play a leadership role in the development of new networking technology. We believe that sophisticated, but easy-to-use laboratory facilities, can help reduce the barriers that can inhibit student entry into engineering studies and can play an important part in reversing the decline in U.S. engineering enrollments, particularly among women and minority students.
Back to News Directory