A model of stream processing of experimental data in distributed systems

Modern experimental stands generate long-lasting and intense data streams. Limited computing resources of stands and increasing the accuracy requirements of algorithms make it impossible to provide a local processing. However, a high bandwidth optical lambda networks allows one to transfer calculations to remote supercomputers. It is extremely important to organize an efficient parallelism in all components of the distributed system. This paper considers a model of stream processing of experimental data and focuses on the use of high-speed fiber-optic long-distance data networks for communication of experimental stands with high-performance computing systems. This work was supported by the Russian Foundation for Basic Research (project 11-07-96001-p-ural-a) and Ural Branch of Russian Academy of Sciences (project 12-П-1-2012). The paper was recommended for publication by the Program Committee of the International Scientific Conference “Scientific services in the Internet: search for new solutions” (http://agora.guru.ru/abrau).