Faculty and staff at the University of California at Santa Cruz are engaged in several very large projects that involve managing and analyzing massive amounts of data. For example, the Astrophysics group led by Prof. Stan Woosley generates 30 TB of data each week that they would like to transfer to the local Hyades cluster for further analysis. The Santa Cruz Institute for Particle Physics is involved in a number of network-intensive research projects, including ATLAS, VERITAS, DES, and LSST, that are currently bandwidth-limited by UCSC’s existing network connections. The Genome Browser (http://genome.ucsc.edu), which is used by researchers around the world for an interactive, graphical display for comparing genomes, handles peak loads that saturate its current 1 Gb/s link. The Cancer Genome Hub project, sponsored by NIH, anticipates transfers of almost 33 PB of data between the CGHub site and researchers from across the country over the three and a half years of the project. The CGHub data repository itself is at SDSC, but the repository developers and operators reside at UCSC and require high performance access to the data for the purposes of development, analysis, and archiving.
The goal of this project is to redesign the research network at UCSC to provide the research projects with the high performance network support they need. The architecture of the network will be flexible and expandable, so as new projects come online, the network can be adapted to handle their needs. The network architecture will provide enhanced measurement and monitoring capabilities necessary for the support of high performance network environment, as well as the ability to deploy novel and experimental technologies (notably, OpenFlow will enable experiments with routing). Further, requested support personnel will act as a bridge between the research projects using this enhanced network and the central campus IT staff managing the network, so researchers can obtain the optimum network performance that they need, and the network itself can be tuned as appropriate to minimize overhead.
These improvements will enable research projects requiring high performance networking to take full advantage of the throughput of high speed networks, such as the CENIC HPR network, connected to the campus border.