High Performance Networking

High-performance networks (HPNs) will play an important role in the success of mission critical, net-centric operations as data volumes and near real-time data processing requirements continue to increase. With high bandwidth, low latency and high reliability, HPNs provide great benefit to organizations that rely on distributed computing environments. For instance, activities like datacenter replication, datacenter disaster recovery, and high-performance distributed computing require high volume data transfer and low network latency to meet critical mission requirements. HPNs with dynamic circuit connection capabilities can make expensive high-performance network resources more accessible and manageable. This paper focuses upon the use of HPNs to enable distributed computing.

Distributed computing allows one or more applications to make use of computing and data resources located at widely separated locations. In theory, this allows users to bring more computing power and better data to bear on their problems, but in practice the difficulties imposed by distributed computing have often outweighed the benefits, resulting in limited acceptance of distributed computing by the user community. A major limitation on distributed computing has been the low bandwidth and high latency associated with conventional wide-area networks.

One of the enabling technologies for widespread adoption of distributed computing in high-performance applications is efficient high bandwidth, low latency networking. With the widespread adoption of dense wavelength division multiplexing (DWDM), networks are beginning to have enough capacity to dedicate bandwidth to individual data flows. These dynamically allocated resources provided distributed computing applications with high bandwidth and low latency while using network resources efficiently. For example, the Internet2 (www.internet2. edu) optical backbone network provides DWDM links with bandwidth capability up to 100Gbps, of which roughly half is available for dynamic circuit reservations. However, using this networking technology effectively and transparently remains a challenge. For this project, Northrop Grumman has built a testbed to investigate hardware and software architectures for HPN. Investigations to date have been in the areas of data transport protocols, control plane APIs, and application frameworks. We have also established partnerships with government agencies and universities to demonstrate our solutions using real-world applications.