We have extensive experience in the development and implementation of distributed and parallel systems, performance modelling and distributed algorithms.
The various computational problems require ever faster computers, and the speed of the technologies used seems to be reaching the speed limits of conventional integrated circuits. To further increase performance, a systems engineering solution is offered, which is embodied in parallelisation.
The use of resources in geographically separated locations provides an additional incentive to use parallel and distributed systems.
The use of clusters to perform general-purpose high-performance computing tasks is also becoming more common as the performance of desktop computers and the speed of network connectivity have increased significantly.
Pyramid cluster system
The Pyramid Cluster System aims to build a parallel programming environment based on a PC cluster.
The Pyramid system is essentially an operating system extension that can link multiple, separate desktop computers into a single virtual computer. The system is based on message handling, and its main goal is to create an efficient and unified cluster programming environment that is well suited to object-oriented languages and allows the use of message handling-based programming methods. An additional important feature of the system is that it not only implements location-independent addressing and messaging, but also provides the possibility for high-level OS-independent management of the GUI and threads. In addition, the Pyramid system provides a laboratory measurement environment for testing various distribution algorithms. These seek to answer the question of which distribution to use for each type of task, depending on the system parameters.
Distributed data mining
The aim of the project is to develop methods and algorithms within the field of data mining that implement traditional data mining tasks in a distributed way. Performance modeling and performance distribution are also important objectives.
Distributed computing and storage systems
Research and development in this theme will support the solution of problems with high computational and/or storage requirements. Distributed systems provide efficient and economical solutions to this challenge through the rapid evolution of network technologies. Significant collaboration has been established in this area with international research groups, notably in the application and development of grid and cloud systems, such as the computer support to the LHC experiment at CERN. In many research areas, such as particle physics, huge amounts of data need to be efficiently stored and managed. Another sub-area of the research theme focuses on distributed file systems such as HP SFS.
