Architecture, methods and tools to build a compute node and small cluster application that can scale with on-demand high-performance computing (HPC) by leveraging the cloud are presented for scaling of video analytics applications. Exotic HPC architectures with custom-scaled processor cores and shared memory interconnection networks are being rapidly replaced by on-demand clusters that leverage off-the-shelf general-purpose vector coprocessors, converged Ethernet at 40 Gbit/s per link or more, and multicore headless servers. These new HPC on-demand cloud resources resemble what has been called warehouse-scale computing, where each node is homogeneous and headless and the focus is on total cost of ownership and power use efficiency overall. However, HPC has unique requirements that go beyond social networks, Web search, and other typical warehouse-scale computing solutions. This series takes an in-depth look at how to address unique challenges while tapping and leveraging the efficiency of warehouse-scale on-demand HPC. The approach allows the architect to build locally for expected workload and to spill over into on-demand cloud HPC for peak loads. Video analytics for high resolution digital video will be demonstrated and have been used to compare system design and to benchmark this more data-driven HPC workload compared to more traditional LINPACK floating point oriented benchmarks and workloads. The testing completed has made use of ARSC Pacman (Arctic Region Supercomputing Center, 2696 compute cores) and the JANUS system (16,416 compute cores) as well as small scale clusters and GPU/FPGA and Multi-core co-processors. Comparative results and strategies for scaling video analytics will be provided.
Dr. Sam Siewert is an assistant professor in the Computer Science and Engineering department at the University of Alaska Anchorage. He is also an adjunct assistant professor at the University of Colorado at Boulder and teaches summer courses in the Electrical, Computer, and Energy Engineering department. As a computer engineer, Dr. Siewert has worked in the aerospace, telecommunications, and storage industries since 1988. Ongoing interests as a researcher and consultant include scalable systems, computer vision, hybrid reconfigurable architecture, and operating systems. Related research interests include real-time theory, digital media, and fundamental computer architecture. He received his B Sc in Aerospace and Mechanical Engineering from University of Notre Dame in 1989, and M Sc (1993) and Ph.D. in Computer Science in 2000 from the University of Colorado at Boulder.