Researchers often generate very large amounts of data in their work, so much so that their data stresses the underlying computer networks. Some academic institutions build a separate network for research data, but this is very expensive. Archipelago is a project that seeks to improve the performance and security of computer networks to make it easier and less costly for scientists to move, store, and analyze their data. It uses relatively small numbers of very fast, flexible network devices which use a technique known as Software Defined Networking (SDN) to steer research data around bottlenecks giving the appearance of a separate science network without having to spend all the money needed to build a true second network.
Archipelago addresses shortcomings that have limited adoption of SDN within campus networks, beginning with evaluation of next-generation SDN forwarding elements, SmartNIC (Smart Network Interface Card) devices and purpose built SDN data planes to create its hybrid architecture. This architecture connects islands of SDN by enforcing traffic policies at nearby distribution points linked to commodity servers with SmartNICs or next-generation hardware-driven SDN forwarding elements (depending on scale) that provide control plane functionality. This shift from software to SmartNIC forwarding accelerates data plane performance and efficient use of distributed CPU resources for the control plane frees server resources for other uses. This robust policy-driven architecture enables fine-grained traffic forwarding policies for microsegmentation, enhanced cybersecurity, efficient data transfer, and other applications.
Utilizing agile software development techniques that are informed by the deployment experience at Duke and within partner environments, this project will deliver a powerful, vendor-agnostic research network design that is easy-to-adopt for institutions that have been reluctant to venture into SDN architectures due to personnel or technical resource constraints. Since the Archipelago architecture augments existing networks incrementally (and therefore cost effectively) and because its software components are deployed via open-source licensing, it will provide new and powerful capability to institutions that already have established SDN networks, as well as for those that have been reluctant to venture into SDN architectures due to financial constraints.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
