 |
||
|
||
Home > TERATEC FORUM > > Workshop 1
TERATEC Forum 2014
HPC & Big data applications and solution architectures: evolution and convergence
Chaired by Pascale BERNIER-BRUNA, BULL - Marie-Christine SAWLEY, INTEL - Jean-Philippe NOMINE, CEA
New challenges are emerging from the expanding region where extreme scale computing and Big Data intersect. Indeed, HPC centres are seeing a sharp increase of incoming data to be analysed and used in the central facilities, a new and unprecedented shift in paradigm for HPC industry. On the other hand, the volume of simulation data to be used for further insight through postprocessing is still growing. Analysis and storage capacities have to be driven by the optimum between efficiency and operational costs in order to scientifically exploit this bi-directional data-flow. The risk of seeing a widening of the gap between capacity/bandwidth and computing capacity as we move towards exascale is real.
The talks of this session will give new perspectives on state-of-the-art integrated solutions based on new workflows, storage and file systems for high capacity, complete platforms and innovative software solutions for deep insight into large volume of complex data.
 Running Hadoop Map Reduce Jobs on LustreGabriele PACIUCCI EMEA Lustre Solution Architect Intel - Zhiqi TAO, Ph. D. ASMO Lustre Solution Architect Intel
|
| Data Storage for the era of Converged Big Data and HPC
Torben KLING-PETERSEN, principal architect, XYRATEX
|
| Fast data meets HPC at TERALAB, dedicated to real-time analytics use cases
Pierre PLEVEN, Direction de l’innovation, INSTITUT MINES TELECOM
|
| Big Data with MATLAB Ascension VIZINHO-COUTRY, MATHWORKS |
| Addressing Open Source Big Data, Hadoop, and MapReduce limitations Emmanuel LECERF, IBM |
|
Abstract & Biography
Download the presentation
© TERATEC 2014 |
Our modern world is facing a new challenge: the flooding of data. Be them coming from scientific experiments or more private usages, those huge amounts of data cannot be handled by classical tools anymore. One needs to use HPC and Business Intelligence large datawarehousing tools, in order to scale innovative methods to manipulate and extract valuable information of such overwhelming databases. The Big Data session is tryed to cast a light on the challenges arising from large volumes and short processing times in relation with HPC and Business Intelligence perspectives. Actors from the scientific and industrial worlds have illustrated those two aspects. They have also explained the relation between HPC and Business Intelligence approaches as a convergence to the Big Data concept.
Scientists and engineers have long been producing and using big datasets, be there numerical simulation results, observed or measured data, or digital models. These data have been tremendously growing in volume while their distribution and circulation on networks has been exacerbated. But we are now also surrounded and overwhelmed by massive flow of business, administrative, financial, multimedia, social network data, as well as data related to sensor networks or mobile objects. Be there public or private, creating meaning and value from these data deluges will require radical improvement and new approaches for their storage, processing, analysis and exploitation. In this session leading observers and players of these “big data” phenomena will share their vision and give us food for thought through a diversity of applications and approaches.