This Small Business Innovation Research proposal presents AutoPHY, an autonomous, virtualized layer 0 interconnection service for data centers and cloud computing. The Open Systems Interconnection (OSI) stack representing the Internet infrastructure consists of layers 0 to 7. Today all layers can be automatically configured and managed through software, with the exception of the physical interconnection layer 0, which is still managed manually. More than 60% of data center outages are now attributed to human errors occurring at layer 0. The AutoPHY service will for the first time automate the management of the millions of physical interconnects in a typical network, at a scale and cost previously impossible. Unique robotic fiber optic cross-connect hardware married with sophisticated Knots and Braids Switching (KBS) algorithms enable a breakthrough switch technology that, for the first time, scales in complexity as N, the number of ports, rather than N2. This technology offers unprecedented capabilities in network provisioning, emergency failover, network partitioning and bandwidth grooming for flexible management of network interconnections at scale, with seamless federation across geographically dispersed, separately managed centers. AutoPHY improves the flexibility and simplicity of cloud network management, secures networks from human error, and enables a scalable, efficient and resilient network infrastructure. The broader impact of this project includes the advancement of standards-based network infrastructure and open source Software-Defined Networking (SDN) networking technologies, and more specifically the development of hardened physical layer management tools and services for deployment in production multi-tenant and mega data centers. The technology to dynamically control and optimize the physical network topology will be used by the AutoPHY global manager, in coordination with SDN controllers, to build high level network services including Automated Smart-Hands and Automated Failover. AutoPHY will reduce cross-connect and provisioning times from days to minutes and eliminate the errors resulting from manual intervention. These developments will be a significant contribution to the same important goals that drive the SDN movement. Whereas current SDN efforts are focused on layers 2 and 3 of the network, AutoPHY enables software programmability to be cost-effectively extended to the physical network topology. By improving the economics and performance of layer 0 at the cloud scale, AutoPHY has the potential to transform cloud computing through higher utilization and support the broader Department of Energy mission.
