Quantum Monte Carlo and Data Management in Grid Middleware
Appears in the following Collection
- Fysisk institutt 
AbstractThe task of this thesis is twofold. First, the thesis considers a genre of physics problems and efficient algorithms for solving them. Second, the thesis considers distributed computing resources and how to connect and utilize them efficiently. While the two tasks at first thought may seem unrelated, each task has been a motivating factor for the other in science in the last half century. Nuclear and particle physics strives to understand how the world is built by exploring its smallest building blocks and how they are connected. This question spawns a large number of computational problems with no upper limits to the amount of computing resources needed to solve them. In many cases advances in physics are limited by the amount of computing resources, thus motivating the work of connecting computing resources to gain more computing power. On the other hand, advances in distributed computing technology are driven by the demand for easily accessible computing power. While the physics community is of course not the only scientific nor non-scientific community with growing needs for computing resources, increasingly large physics experiments have generated demands that would have been considered impossible by smaller experiments. As a leading example, the European Organization for Nuclear Research (CERN) created the world wide web as a means to simplify communication and collaboration between scientists and is now in the forefront of developing distributed computing as a consequence of the computing needs of the largest physics experiment ever built, the Large Hadron Collider. The two main topics of this thesis, Quantum Monte Carlo and grid middleware, are two components needed to achieve the same goal. The numerical method of Monte Carlo provides a means to study physical problems in an ab initio fashion, applying basic physical rules to simulate physical experiments. The grid middleware provides global access to the computational resources needed to simulate the experiments. The goal is to improve our understanding of physical phenomena by means of computer simulations. While the physical phenomenon in focus for this thesis, the physics of Bose-Einstein condensates, is described in the articles in Appendices A.1 and A.2, we will in the remainder of this introductory chapter introduce the method of Quantum Monte Carlo in Section 1.1 and the grid technology in Section 1.2, before presenting the outline of the thesis in Section 1.3.
List of papers
Paper 1 Vortices in atomic Bose-Einstein condensates in the large-gas-parameter region. J. K. Nilsen, J. Mur-Petit, M. Guilleumas, M. Hjorth-Jensen, and A. Polls. PHYSICAL REVIEW A 71, 053610 2005 DOI: 10.1103/PhysRevA.71.053610
Paper 2 MontePython: Implementing Quantum Monte Carlo using Python. Jon Kristian Nilsen. Computer Physics Communications 177 (2007) 799–814 DOI: 10.1016/j.cpc.2007.06.013
Paper 3 Simplifying Parallelization of Scientific Codes by a Function-Centric Approach in Python. Jon K. Nilsen, Xing Cai, Bjørn Høyland and Hans Petter Langtangen. Submitted to: Computational Science & Discovery
Paper 4 Recent ARC developments: through modularity to interoperability. O Smirnova, D Cameron, P Dóbé, M Ellert, T Frågåt, M Grønager, D Johansson, J Jönemo, J Kleist, M Kocan, A Konstantinov, B Kónya, I Márton, S Möller, B Mohn, Zs Nagy, J K Nilsen, F Ould Saada, W Qiang, A Read, P Rosendahl, G Röczei, M Savko, M Skou Andersen, P Stefán, F Szalai, A Taga, S Z Toor and A Wäänänen. Article submitted to Journal of Physics, Conference Series. DOI: 10.1088/1742-6596/219/6/062027
Paper 5 Chelonia – A Self-healing Storage Cloud. Jon K. Nilsen, Salman Toor, Zsombor Nagy and Bjarte Mohn. Article to appear in the proceedings for Cracow ’09 Grid Workshop, Oct. 12-14.
Paper 6 Article submitted to Journal of Parallel and Distributed Computing. Performance and Stability of the Chelonia Storage Cloud J. K. Nilsen, S. Toor, Zs. Nagy, B. Mohn, A. L. Read
Paper 7 Parallel Monte Carlo simulations on an ARC-enabled computing grid. Jon K. Nilsena, Bjørn H. Samseta. Article submitted to Computer Physics Communication.