The paper presents an application of modern computer services known as cloud computing for the simple coil geometry optimization problem. The Monte Carlo method is known for its robustness, but also low convergence. The latter shortcoming could be eliminated by large and affordable computational power offered today by cloud providers. The described architecture of the simulation system is based on Microsoft Azure platform with HTCondor as a job manager.
Słowa kluczowe: cloud computing, optimal design, Monte Carlo method
Artykuł przedstawia wykorzystanie usług obliczeniowych na przykładzie prostego zagadnienia optymalizacji kształtu cewki. Metoda Monte Carlo jest znana ze swojej skuteczno´sci, a jednocze´snie z bardzo niskiej zbie˙zno´sci. Wad˛e t ˛ a mo˙zna skutecznie ograniczy´c poprzez wykorzystaniem du˙zych i tanich mocy obliczeniowych oferowanych dzisiaj przez dostawców usług ’chmurowych’ (ang. cloud computing). Opisana architektura systemu symulacyjnego oparta jest na platformie Microsoft Azure oraz zarza˛dcy zadan´ HTCondor.
Keywords: obliczenia rozproszone, optymalizacja, metoda Monte Carlo
An important trend in modern computational science is certainly cloud computing. Although this term has been created by the marketing departments in early 2000’s, and a history of distributed and parallel processing in computer science is much longer, one could realize novelty in cloud computing approach. Utilization of commercially managed resources for scientific tasks has significant advantages, such as a flexibility to use different hardware, no need for infrastructure investments. On the other hand, important challenges are raised  and it is known that not all types the computing problems benefit from a loosely coupled architecture in the same way. Groups of problems, which could be easily transferred into the cloud infrastructure are independent simulations, sometimes called as an ’embarrassingly parallel’ problems. Sensitivity analysis , stochastic simulations  are just examples of the problems which require a large number of simulations. Another are DNA alignment in bioinformatics, 3D scene rendering in computer graphics or Monte Carlo methods, which are the main subject of this article. The Monte Carlo methods are based on probing parameters space with the use of a random generator. Applications of such simple but robust solution are wide, and they are especially compatible with the structure of the cloud services . Stochastic optimization using Monte Carlo sampling is one of them . In this paper, stochastic optimization technique is used to design the magnetic coil system. The developed simulation platform is constructed using HTCondor embedded in the Microsoft Azure environment as presented in Fig. 1. Finite element method solver has been constructed using FEniCS library as described later in the paper. Coil design problem is a test case, which could be also successfully solved using gradient optimization techniques. However it should be treated as a benchmarking tool to study e [...]
 Angelo G., Marzolla M., New trends in parallel and distributed simulation: From many-cores to Cloud Computing, Simul. Model. Pract. Theory, 126 (2014)  Krupa A., Sawicki B., High-resolution scatter analyse using cloud computing, Przeglad Elektrotechniczny, 91 (2015), Nr 12, pp. 140-142  Krupa A., Sawicki B., Measurement-based stochastic models of biological materials, 18th International Conference on Computational Problems of Electrical Engineering (CPEE 2017), Kutna Hora, 11-13th September 2017  Wang H., Ma Y., Pratx G., Xing L., Toward real-time Monte Carlo simulation using a commercial cloud computing infrastructure, Phys. Med. Biol., 56 (2011)  Homem-de-Mello T., Bayraksan G., Monte Carlo sampling-based methods for stochastic optimization, Surveys in Operations Research and Management Science, 19 (2014)  Krupa A., Sawicki B., Massive Simulations Using MapReduce Model, IAPGOS, 5 (2015), No 4, pp. 45-47  Thain D., Tannenbaum T., Livny M., Distributed computing in practice: the Condor experience, Concurrency - Practice and Experience, 17 (2005), pp. 323-356  Srirangam R., Patil R., Azure GAHP Server for HTCondor, Microsoft, (2017)  Rynge M., Desinghu B., Thapa S., Harstad E., Large Scale Computation with HTCondor, SWC-OSG Workshop materials, (2017)  Alnæs M. S., Blechta J., Hake J., et. al., The FEniCS Project Version 1.5, Archive of Numerical Software, 3 (2015), No. 100, pp. 9-23  Yamada S., Ikehata Y., Ueno T., Nagae H., Wireless Power Transfer System for Hyperthermia Therapy, Proceedings of the 2nd Frontiers in Biomedical Devices Conference, (2007)