ICONS home > Project Documents > D19
Of this deliverable, the full document (in pdf, 1,97 MB) is also available.
The primary objective of distribution of the ICONS platform components, and consequently the ICONS knowledge management application components, is to achieve system scalability. We propose to exploit the inherent load balancing and failover potential of the J2EE application servers as well as that of data distribution supported by SDDS.
The distributed ICONS architecture comprises software components to be executed within containers managed by the J2EE application servers, namely the Web Container and the EJB Container, the software components to be managed by JADE (Java Agent Development Environment), and ICONS Services processing the asynchronous elementary operations implemented in the system. In the latter case we develop a service scheduling algorithm providing the required load balancing features based on the Petri Net representation of the application workload. The ICONS Services are implemented on the basis of the SOAP standard according to the Web Service architecture. Due to their internal nature, we are not providing the UDDI and the WSDL features. The distribution goal is to provide for the uniform distribution of utilisation of the underlying hardware architecture comprising processing service centres and data storage service centres. The optimisation goal function is minimising the composite standard deviation of utilisation of the above two processing centre subsets under the memory capacity and utilisation level constraints.
A crucial aspect of the distributed component architecture is that of the data management components. ICONS load optimization techniques for this part of its architecture rely on the extensive study of the state-of-the-art. After the discussion of the organization and optimization of the processing component, we present in depth this study and its impact on the ICONS system design.