Award Finalist/Winner

SC Conference - Activity Details

Using Phoebus for Optimized Data Dissemination

Martin Swany  (University of Delaware)
Ezra Kissel  (University of Delaware)
Patrick McClory  (University of Delaware)
Aaron Brown  (Internet2)
Challenges Session
SC08 Bandwidth Challenge: Unleashing the Power of the Network
Tuesday,  02:45PM - 03:00PM
Room 17A/17B
Scientific computing involves working with large, distributed datasets. This requires effective use of both available processing capability and the means to optimize the transfer of actual data across remote compute nodes. Phoebus is an architecture that increases network throughput by intelligently segmenting and routing across heterogeneous paths. Phoebus can make decisions about network optimization based on network information provided by the perfSONAR framework. The analysis and summarization of available network performance data can itself be a compute-intensive task. For this challenge, we intend to use distributed computing resources to analyze network data. The results of this analysis will be fed into routing models. This will initially require massive distribution of data and will use a model similar to that planned for Large Hadron Collider data, namely a hierarchy of dissemination and analysis.
   IEEE Computer Society  /  ACM     2 0   Y E A R S   -   U N L E A S H I N G   T H E   P O W E R   O F   H P C