Instituto Tecnológico de Canarias - Atlante Node
Atlante supercomputer joined the RES on February 16th 2009, becoming its 8th member and the 2nd member from the Canary Islands. It is managed by Instituto Tecnológico de Canarias (ITC), a public company of the Canary Islands Regional Government, that promotes the industrial development of the region, fostering research, development and innovation in emerging technological fields, in close collaboration with companies and research institutions.
The Atlante node is located at the Science and Technology Park of the University of Las Palmas de Gran Canaria . The cluster is formed by 84 IBM JS21 blade servers with dual core PowerPC 970MP processors and 8GB RAM (336 CPUs in total), reaching 3.36 TFLOP/s. Atlante uses a Myrinet-2000 interconnection network and offers 8TB of storage disk.
The technical staff of Atlante comprises two engineers from ITC (Yeray Gutiérrez Cedrés and María Belén Esteban Sánchez), who are responsible for system management and user support, and a group manager (Juan José Ascanio Amigó). A local Access Committee allocates the 80% of local CPU time amongst users from Canary Islands's Government, private companies and R&D groups, while the remaining processing time is provided to the RES network.
Technical and Scientific Highlights 2011
In 2011 Atlante node executed applications corresponding to Canary islands's Companies and Researchers with more than 2 million hours of CPU time, related to the following projects:
Key Publications 2011
- L. Cana, D. Grisolía-Santos, L. Álvarez. "First results on the modeling of the Gran Canaria atmospheric wake". Geophysical Research Abstracts, Vol. 13, 10304, 2011. General Assembly 2011. European Geophysical Union.
- L. Cana, D. Grisolía-Santos, L. Álvarez. "Local flow under weak trade wind flow regime". General Assembly 2011. European Geophysical Union. Geophysical Research Abstracts, Vol. 13, 10328, 2011
- Troupin. C., E. Mason, J.M. Beckers, P. Sangrà, "Generation of the Cape Ghir upwelling filament: A numerical study". Ocean Modelling, 41,1-15
- Enrique Fernández-Perdomo, Jorge Cabrera-Gámez, Daniel Hernández-Sosa, Josep Isern-González, Antonio C. Domínguez-Brito, Víctor Prieto-Marañón, and Antonio G. Ramos. Adaptive Bearing Sampling for a Constant-Time Surfacing A* path planning algorithm for gliders. In Proceedings of the IEEE 2011 International Conference on Robotics and Automation (ICRA 2011), Shanghai, China, May 2011.
- Josep Isern-González, Daniel Hernández-Sosa, Enrique Fernández-Perdomo, Jorge Cabrera-Gámez, Antonio C. Domínguez-Brito, and Víctor Prieto-Marañón. Path planning for underwater gliders using iterative optimization. In Proceedings of the IEEE 2011 International Conference on Robotics and Automation (ICRA 2011), Shanghai, China, May 2011
- Pérez, J.C., García‐Lorenzo, B., Díaz, J.P., González, A., Expósito, F., Insausti, M. Forecasting precipitable water vapour at the Roque de los Muchachos Observatory. M. Proceedings of SPIE ‐ Volume 7733, Ground‐Based and Airborne Telescopes III; San Diego, CA; 27 June 2010 ‐ 2 July 2010. ISSN: 0277786X, ISBN: 978‐081948223‐5, DOI: 10.1117/12.859453.
- Antonio Eff‐Darwich, Juan C. Pérez, José Fernández, Begoña García‐Lorenzo, Albano González and Pablo J. González. Using a Mesoscale Meteorological Model to Reduce the Effect of Tropospheric Water Vapour from DInSAR Data: A Case Study for the Island of Tenerife, Canary Islands. Pure and Applied Geophysics. 2011, DOI: 10.1007/s00024‐011‐0401‐4.
- Juan Mendez. Cooperating Multi-Core and Multi-GPU in the Computation of the Multidimensional Voronoi Adjacency in Machine Learning Datasets. International Conference on Parallel and Distributed Processing Techniques and Applications (PDPTA'10), Las Vegas, USA
- Charles Troupin. Study of the Cape Ghir upwelling filament using variational data analysis and a regional numerical model. Université de Lièje (Belgium). Supervised by Jean Marie Beckers and Pablo Sangrà.
Key Projects 2011
- PHOTOVOLTAIC SYSTEMS SIMULATOR (Simulation of Photovoltaic Energy in Gran Canaria)
The project deals with meteorological data processing to estimate the amount of energy that Photovoltaic Systems allocated in Canary Islands would be able to generate according to the expected weather conditions, which are estimated using Numerical Weather Prediction models (MM5) executed by the Atlante supercomputer. The Photovoltaic Systems Simulator was run in Atlante for the whole of 2011, and its estimations are used for the power load dispatch center operations.
- DEVELOPMENT OF AN ENERGY FORECAST SYSTEM WITH APPLICATIONS IN PHOTOVOLTAIC PLANTS BY USING NUMERICAL MODELS
The technical complexity of energy systems and the increasing tendency of introducing renewable energies into these systems, makes it essential to use forecasting techniques to try to predict solar radiation. This makes it possible to organize and optimise all the net's resources. Therefore, the aim of this project is to develop a system which can predict the correct amount of solar energy which is going to reach the Earth’s surface at the Canary Islands. The numerical model used was the Weather Research and Forecast model (WRF), executed in Atlante.
- METEOSAT SECOND GENERATION SURFACE TEMPERATURE ASSIMILATION FOR WRF MODEL OVER CANARY ISLANDS DOMAIN
One of the problems introducing electricity generated by renewable-based generation units is that the amount of energy generated depends on the weather conditions. In order to balance generation/demand and also guarantee the stability oF the electrical system, the electrical network manager needs to know the clean generation forecast. The developmental tendency and primary method to deal with this problem involves predicting wind and/or solar radiation fields using numerical Weather Prediction Model (NWP). In this study Weather Research and Forecasting (WRF-ARW) model was used. The objective of this study is improve the initial ST conditions introducing values from TIR channels, calculated by MSG satellite, in order to obtain more accurate results using WRF model for solar shortwave flux at surface field.
- ADAPTIVE EVOLUTIONARY TRAVEL ROUTE SYSTEM (AETROS)
The Adaptive Evolutionary Travel Route System (AETROS) model was designed to solve the travel route assignment problem in a new, dynamic way. Urban traffic present highly changing scenarios where statistic based optimisation approaches cannot cope with unforeseeable situations. AETROS is designed to be constantly adapting to the current situation and drivers are not assigned a fixed origin-destination route. Instead, as they approach an intersection they will receive the latest optimal option for their destination.
- DRAGON AIRBORNE POWER
Dragon Airborne Power (formerly Persan Engineering) is actively pursuing the development of high-altitude wind power systems, which calls for the computing-intensive simulation of the external aerodynamic flow and energy-capturing process on a proposed airborne turbine model using the finite-volume discretization of the Navier-Stokes equations. To this end, the Code Saturne flow solver was extensively used on Atlante, leading to the assesment of power levels, lift/drag/moment coefficients and stability derivatives needed to design a working prototype in the short term future.
- SIMULATION STUDIES AND DESIGN OF MATERIALS AND NANOMATERIALS UNDER EXTREME CONDITIONS
One of the most fundamental goals in high pressure research is to mimic processes and phenomena similar to those occurring in the interior of the Earth and other planetary objects, and to understand physical and chemical properties that appear in exotic phase after phase transitions under extreme conditions. This project aims to develop an ab initio study of the structural, electronic and dynamical properties under extreme conditions, combined with different extreme conditions experimental techniques performed in synchrotron and neutron sources.
The problem of Simultaneous Localization and Mapping (SLAM) addresses the issue of a mobile robot moving through an environment of which no map is available a priori. The goal of SLAM is to reconstruct a map of the world and the path taken by the robot, being a key prerequisite to truly autonomous robots. The work focuses in multi-sensor fusion to improve the quality of the robot pose estimation as well as the map of the surrounding environment. Currently, we are also in the process of integrating inertial motion information to aid the state estimation. More specifically, we want these techniques to perform in real time, in which the quality of the map might be penalized, but enable us for active perception. Still, high computational task are required to obtain a ground truth, for which we process collected data offline in batch mode.
- EFFICIENT COMPUTATION OF VORONOI NEIGHBORS
Some algorithms in Patter Recognition and Machine Learning as neighborhood-based classification and dataset condensation can be improved with the use of Voronoi tessellation. We study the use of a cooperative framework where multiple Cores in host and multiple GPUs cooperate to compute the Voronoi adjacency relationship in multidimensional Machine Learning datasets.