SCL at SEE-GRID-SCI User Forum 2009

The first User Forum 2009 conference of the SEE-GRID-SCI project was held in Istanbul, Turkey on 9 and 10 December 2009. On behalf of the project, the conference was organized by the Turkish Academic Network, Middle East Technical University and Bogazici University. Main focus of this event was presentation of scientific results obtained using distributed Grid e-Infrastructures and high performance computing resources, as well as on the distributed and parallel computing algorithms and implementations. During the talks and poster sessions various aspects of Grid and high performance computing were presented, including the newly developed applications deployed by research institutes and universities from the South Easter Europe region. This event also hosted several invited talks by speaker representing other e-Infrastrucuters and initiatives, including European Eart Science collaboration, Asian and Latin American Grid projects, thus making the meeting an excellent opportunity for networking of different communities.

SCL contributed to this event with three papers presented in talks by Vladimir Slavnic:

- Operational Grid tools developed at SCL
A number of tools are used in everyday Grid operations and operational support by site administrators in order to allow efficient and reliable management of Grid services and their full availability to end users. Such tools perform various functions, from monitoring and problem diagnostics to site installation, configuration and management. In the Scientific Computing Laboratory (SCL) of the Institute of Physics Belgrade we have developed several tools useful maintenance and management of gLite-based Grid sites. Some of these tools can be also used on general purpose Linux-type clusters. Here we report on the SCL tools most used in everyday operations.

- Grid Site Monitoring tools developed and used at SCL
In today's deployment of Grid services, continuous monitoring of computing and storage resources and core services is crucial for achieving high level of service reliability. A number of system parameters has to be under constant observation and control: from hardware conditions (temperature of the motherboard and of the CPU, fan speed, disk state) through state of operating system, to Grid-related events and daemons. In the Scientific Computing Laboratory of the Institute of Physics Belgrade we are using several such monitoring tools, with a number of them partially or fully developed and publicly available. Here we present the most used Grid monitoring tools at SCL.

- Optimization and Porting of the Path Integral Monte Carlo SPEEDUP Code to New Computing Architectures
Many complex quantum physical systems can be most effectively described by the path integral formalism. The SPEEDUP code implements recently introduced analytical approach that systematically improves convergence of numerically calculated transition amplitudes of a generic quantum non-relativistic theory that leads to significant speedup of Path Integral Monte Carlo algorithms. Optimization, porting and testing of the SPEEDUP code was done on several new computing architectures: IBM POWER6, PowerXCell and the latest Intel CPUs. We give overview of optimization techniques used and present results on the use of advanced features of new CPU types, which can be efficiently applied for the optimization of the SPEEDUP code.

  • user_forum-001
  • user_forum-002
  • user_forum-003