Session 6

Char: Kate Keahey (ANL), Gabriel Antoniu (INRIA), Rosa Badia (BSC)
Speakers:
In addition to the chairs, we will have talks from Ramon Nou (BSC), Bruno Raffin (INRIA). We will also solicit additional speakers from these institutions as well as from NCSA.

Unprecedented growth of opportunities in experimental sciences, together with on-demand access, control over com- puting environment provided by containers and virtualization, and streamlined data processing techniques from the cloud world are revolutionizing the ways in which we can do science. They also however raise questions about how these new datacenter management and programming techniques relate to traditional HPC. Those questions introduce the potential for innovation across the stack including technology for node management (e.g., containers versus virtualization, integration of smart NICs and accelerators), storage (e.g., explicit QoS in storage, exploiting storage hierarchies), networking (SDN, network reservations), resource management (e.g., managing dynamicity, multi-aspect leases), models (improving predictability versus embracing unpredictability, ability to understand and express trade-offs), frameworks (emphasis on loosely coupled programming models, workflows, approaches to re- silience, programmable platforms, malleable applications), data (the emergence of data-centric abstractions, data consistency models). We propose to organize a breakout session to organize and discuss specific ideas and formulate collaborations and partnerships on the convergence of cloud, BigData and HPC technologies across the software stack. The session will be structured around presentation foci in each of the areas described above and will seek to chart out both present and future collaboration on innovative topics.