InfoBeyond is award by Department of Energy to improve perfSONAR-based network anomaly detection
High performance computing (HPC) addresses these computational challenges through the application of parallelization and distribution algorithms or techniques to connected computing units, to perform more complex tasks in a faster manner than a single unit could do alone. For such a purpose, edge computing networks and communications technologies support massive information transmission over significant distance via distributed and networking servers for collaborative computing.
In supporting HPC data communications, perfSONAR is a widely-deployed test and measurement infrastructure that is utilized by science networks and facilities around the world to monitor and ensure network performance. The task of InfoBeyond in this award work is to investigate a new perfSONAR-based approach to improve the network monitoring with a capability to not only detect the network performance anomaly but also fast locate the source of the degradation of the performance.
Such a research is meaningful for a variety of big data applications. The Big data analyzing innovations in transmission, switching, processing, analyzing, and retrieval of information are vital for the success of a wide range of emerging communications technologies, including connected sensors and Internet of things devices, smart grid, smart cities, software-defined networks, network function virtualization, data-driven cognitive networking, cyber security, green communications, etc. The class of computational problems that need to be tackled, such as massive data analysis, pose new challenges in design or develop the above emerging communications technologies.
InfoBeyond studied network performance ranging from multi-domain HPC performance diagnosis by observing the end-to-end throughput, delay, and other metrics to access the anomaly. In this project, InfoBeyond develop novel approaches by exploiting spatial/temporal/topological correlation in network measurements. The outcome of the research could be beneficial for DoE energy-science network (ESnet), global research and education (R&E) network, and other HPC networked systems.