Integrated management framework for dynamic virtual organisations

The trend towards Service Oriented Architecture (SOA) based distributed applications for all kind of business applications lead to a corresponding change in the Grid community. This move away from proprietary solutions and protocols for Grid applications utilizing also Web Services specifications following the SOA paradigm requires also a change of the underlying collaboration model. This model referred to as Virtual Organisation (VO) [1]was originally an almost static collaboration of rather homogeneous resource providers (such as High Performance Computing centres as realised in Uniform Access over the Internet to Computing Resources (UNICORE) [2] or Distributed European Infrastructure for Supercomputer Applications (DEISA) [3]). About ten years earlier the concept of Virtual Enterprises, Collaboration Networks, Alliances, Networks,. . . emerged in the economics research community [4, 5, 6, 7, 8, 9]. These concepts were not limited to the long term oriented goals associated to Grid VOs such as the support of the Large Hadron Collider experiment but targeted at a fast and efficient way to respond on market opportunities. The starting point for this thesis was the development of a more dynamic and business oriented model for IT oriented Virtual Organisations realised by extending the rather limited understanding of Grid VOs with the dynamism of the economic models and introducing the concept of potentially dynamically agreed Service Level Agreements defining the relationship across providers and between consumer and provider(s). Beside this new model allows giving up the assumption that all VO participants agree on the overall goals of the global VO. The SLA based model allows each provider to be driven by their own business objectives and operate autonomously and independent from each other. It is also possible to provide assets to several, potentially competing, VOs at the same time. Using this SLA driven dynamic VO model several business Grid scenarios had been analysed and classified. The classification of the scenarios led to the following abstractions: • Core Service Provider: Potentially large number of providers offering quite raw services very tightly coupled to a physical resource such as data storage or computation. The offered services of the different providers can be easily compared as the number of different services is very limited. • Aggregated Service Provider: Compared to the Core Service Provider the provided services are not tightly mapped on a particular physical resource but combine several core and other aggregated services to a complex, aggregated service. • Service Provider Collaboration: This scenario could be built from a combination of core and aggregated service providers realising a collaboration across them e.g. with specialised roles. So one provider could be specialised on data storage, another one on providing computational services while others are realising an interface to a product database.