Without high-performance networking big data and high performance computing have very limited value. It has long been recognized that the highest performance data transfers should directly run the on top of optical circuits without Ethernet switching or IP routing. This direct layering avoids switching-related packet loss and provides a cost effective service. Optical equipment now exists that will permit the dynamic establishment of single application light paths. However, the time required to set up optical circuits is too long for short or small transfers ? taking a second to set up a 100 Gigabit/sec link in order to transfer a small file does not make sense. This project is building an integrated campus network where small data transfers (megabytes or gigabytes) ride conventional networks and big data transfers (terabytes or petabytes) ride directly on optical circuits.
The project is building PROnet, a campus network at the University of Texas, Dallas, which provides the lowest possible network latency (signal propagation only), scalability to 100Gbps and beyond, and equipment having low power consumption and complexity. A user-friendly interface will be developed and deployed for the network inexperienced researcher to request the provisioning of as needed optical. A PROnet controller will be developed for the scheduling and automatic provisioning of optical circuits, which will shorten the circuit reservation procedure time significantly by removing human intervention from the entire process. For inter-domain operation, the open platform OSCARS will be extended to work with PROnet. With this OSCARS extension, provisioning of hybrid (optical and virtual) circuits spanning across multiple (optical and non-optical) domains will be a reality. The software will support secure use of the network via InCommon identities.
