Grid Problem Solving Environments
Grid systems, which are a result of natural evolution of distributed computer systems, enable sharing of geographically-distributed computer resources. This new technology addresses a new kind of a user - virtual organizations which may be spread over many real institutions. The research team from the Institute of Computer Science AGH (Marian Bubak, Włodzimierz Funika, Bartosz Baliś, Maciej Malawski, Katarzyna Rycerz, Marcin Smętek) and ACC CYFRONET AGH (Tomasz Gubała, Piotr Nowakowski, Marcin Radecki, Tomasz Szepieniec) has contributed to this area by developing modeling methods and problem solving environments as well as selected services/components for the underlying middleware. Our recent achievements include:
- a proposal of a unified architecture of Grid services and a tool environment for running interactive compute- and data-intensive applications; this architecture has been implemented in the CrossGrid international testbed,
- elaboration of a tool for online, interactive performance analysis (G-PM) combined with online monitoring of Grid applications (OCM-G). Use of the Grid for interactive applications imposes new requirements on tools supporting performance analysis, namely: online mode of operation and the ability to measure user-defined, application-specific metrics. Furthermore, the monitoring system should be responsible for run-time instrumentation of applications, providing basic measurement mechanisms,
- development of a P2P system for distributed computing based on components (MOCCA) and for monitoring of resources and applications (GEMINI),
- elaboration of a methodology for porting legacy code to Grid systems based on modern service-oriented architectures (the LGF system),
- an integrated framework (G-HLAM) for running High Level Architecture (HLA) distributed large-scale simulations on the Grid,
- a solution for the problem of abstract workflow composition (WCT) in a semantic Grid environment,
- a concept and prototype of a distributed, scalable, semantic-based and Grid-enabled registry for storing information about Grid services, equipped with fault tolerance and data synchronization mechanisms,
- a new approach to monitoring of distributed Java applications with implementation of an online mionitoring infrastructure (J-OCM), which provides uniform support for various tools (debuggers, performance analyzers).
The above research was conducted in collaboration with Universiteit van Amsterdam (Prof. Peter M.A. Sloot), Universitaet Wien (Prof. Peter Brezany), Universitaet Siegen (Prof. Roland Wismueller), Emory University in Atlanta (Prof. Vaidy Sunderam), Universitaet Linz (Prof. Dieter Kranzlmueller), CERN (Dr. Fabrizio Gagliardi), Universitaet Innsbruck (Prof. Thomas Fahringer) and with a number of EU IST projects: CrossGrid, EGEE, K-WfGrid, CoreGRID) as well as projects funded by Polish grants (KBN).
Our main challenge is the elaboration of methods and tools for the development of collaborative large-scale scientific applications running in service-oriented, component, and peer-to-peer Grid systems, based on semantic discovery of services and application composition.