Network-aware Data Management
IEEE/ACM International Conference for High Performance Computing, Networking, Storage and Analysis.
Sunday, November 16, 2014 (NDM'14 has been moved from 11/21/2014 to 11/16/2014)
( 1:30 pm - 5:30 pm )
Ernest N. Morial Convention Center
New Orleans, Louisiana
Journal Special Issue: Corresponding authors of the accepted NDM’14 authors have been invited to submit their extended papers to a special issue in Future Generation Computer System Journal (FGCS, Elsevier – impact factor 2.6). We have sent invitations to corresponding authors of NDM’14 papers in December 2014. Please note that extended versions of workshop papers will undergo another review process for the journal publication. Submission deadline is in February 2015.
Abstract: The landscape of cyberinfrastructure for research is rapidly changing. There is a move towards virtualized and programmable infrastructure. The cloud paradigm enables applications to use computing resources at different places and optimize workflows in either bringing computing to the data or the other way around. Programmable networks allow networks to be utilized in unprecedented ways to create application specific Internets. This talk presents the latest developments in the Research and Education Networks to support Big Data sciences. [slides]
RDMA in the Cloud: Enabling high-bandwidth, low-latency communication in virtualized environments for HPC, Josh Simons, VMware Inc. Office of the CTO
Abstract: While throughput and embarrassingly parallel HPC applications run with very little (<5%) performance degradation in modern virtualized environments, very latency sensitive applications present more of a challenge due to I/O overheads introduced by virtualization. Consequently, enabling use of RDMA-capable interconnects like InfiniBand and RoCE is an important aspect of broadening acceptance and use of cloud computing in HPC. This talk will present our latest performance results using InfiniBand and RoCE in passthrough mode as well as our early experiences with InfiniBand SR-IOV, an important feature that enables multi-VM access to RDMA-connected parallel file systems like Lustre and GPFS. [slides]
NDM'14 proceedings ISBN 978-1-4799-7019-3
5 papers accepted out of 11 submissions
Adaptation and Policy-Based Resource Allocation for Efficient Bulk Data Transfers in High Performance Computing Environments, Ann L. Chervenak (USC/ISI), Alex Sim (LBNL), Junmin Gu (LBNL), Robert Schuler (USC/ISI) and Nandan Hirpathak (USC)
Flexible Scheduling and Control of Bandwidth and In-transit Services for End-to-End Application Workflows, Mehmet Fatih Aktas (Rutgers University), Georgiana Haldeman (Rutgers University) and Manish Parashar (Rutgers University)
Towards Energy Awareness in Hadoop, Krish K.R. (Virginia Tech), M. Safdar Iqbal (Virginia Tech), M. Mustafa Rafique (IBM Research Ireland) and Ali R. Butt (Virginia Tech)
Towards Managed Terabit/s Scientific Data Flows, Artur Barczyk (Caltech), Azher Mughal (Caltech), Harvey Newman (Caltech), Iosif Legrand (Caltech), Michael Bredel (Caltech), Ramiro Voicu (Caltech), Vlad Lapadatescu (Caltech) and Tony Wildish (Princeton University)
NDM'14: IEEE conference record number: 35183
Please note: NDM'14 has been moved to Sunday Nov 16, 2014 from Friday Nov 21, 2014.
SC'14 Website: http://sc14.supercomputing.org/