The grid computing paradigm has facilitated the instrumentation of complex, highly-demanding collaborative applications. The technologies that make grid computing possible have mostly evolved from parallel and cluster systems. Although this has certainly empowered the grid computing field, part of the heritage has been the perception that required network resources are taken for granted. This is precarious, considering that most grids rely on public IP networks, like the Internet, as the underlying network. This assumption has obstructed the path of grid computing. This thesis aims to improve the performance of grid applications by facilitating network-aware grid scheduling. This is achieved by providing network performance information to grid schedulers, allowing them to adapt to changes in the network. The contribution of this thesis is twofold: a novel approach to network measurement that is particularly suitable for grid environments; and a distributed system that collects and manages these measurements, predicts future network performance, and disseminates this information to schedulers. The accuracy and effectiveness of this system is evaluated on a production grid infrastructure used for e-science applications. The outcomes of this evaluation provide a strong argument for the introduction of network-aware grid schedulers, information systems, and job and resource description standards.

Context-awareness is a computer science concept that aims to enhance the usability of computing systems for users by unobtrusively providing services based on user and environmental context. Experimental context- aware systems typically constitute the functions of data acquisition, reasoning, context representation, context storage and context dissemination. Due to the mobile and spatial nature of the computational actors context-aware systems are usually designed with distributed software components. This distribution presents challenges in context provisioning, that is. the coordination and communication of contextual information between distributed components of the system. One of the main challenges is the diversity of settings in which context-awareness functions take place, highlighted by the heterogeneity of devices and the fluid pattern of human behaviour. Moreover, as mobile devices increasingly become the primary tool of user interaction with the environment, the complexity to adequately support awareness for diverse context applications, poses a crucial challenge in development of a large-scale context provisioning systems .. Many prototype context-aware systems have been developed that showcase context-awareness in one application domain or other but large- scale context provisioning, which also facilitates the emerging cognitive role of mobile devices, has proved elusive so far due to multi-faceted challenges in this area. Most of the research in the domain of context-aware computing focuses on context creation, prototyping context-aware applications and context use, while the challenges posed by large-scale context provisioning in such systems have not been adequately addressed. The common theme in existing efforts into developing context-aware systems has been the centralisation of the context management component that acquires, processes and distributes context information. In order to seamlessly coordinate and disseminate contextual information in a large-scale ubiquitous computing environment, a context provisioning approach is required that can overcome scalability, heterogeneity and coordination related challenges. These challenges are addressed in this work and a distributed context provisioning model is presented that can 1) disseminate multi-domain context over a large (geographic or network) scale, 2) provide coordination of context between distributed con- text producing and consuming components, 3) scale well with the increase in the system load due to context related traffic and 4) facilitate user and device mobility within the distributed system. The large-scale context provisioning model proposed by this work, entitled the Context Provisioning Architecture, is based on a federation of context brokers and uses a publish/subscribe based context coordination and communication mechanism. This thesis presents the theoretical modelling and practical implementation of the inter-broker routing of context subscriptions and notifications, management of component mobility, context modelling and context caching. Our proposed federated broker model is the first practical demonstration of federated context brokers for large-scale context provisioning. The load scalability aspect of our proposed model is experimentally evaluated against that of a centralised broker based model and the results demonstrate the improved scalability of the federated broker model. This work also proposes and demonstrates the concept of a context brokering component to manage and facilitate the evolving role of mobile devices towards cognitive context-aware devices.. The novel Mobile Context Broker component reduces the functional burden on mobile device based context consuming/providing applications. Our empirical evaluation shows a significant reduction in average energy consumption on mobile devices during the execution of context-awareness related functions in presence of the Mobile Context Broker. A minor contribution of this thesis is the use and comparison of cache replacement policies that can be utilised in a context caching component of the system. Our experiments reveal that caching is a viable strategy for improving the scalability of a context provisioning system and different cache replacement policies can be utilised within this strategy. Furthermore, this work also demonstrates a unified context model that is used not only for context representation but also for context coordination across distributed context brokers and context consuming/providing applications. Collectively, these contributions provide a holistic architectural level support and conceptual foundation that can serve as a guideline for designing large-scale context provisioning systems.

The rapid growth in demand for computational and data resources have created the need for efficient ways to manage the process of describing and discovering these resources. The most known technology for resource sharing is Grid computing in which multiple computing nodes share computational resources to provide services. However, when investigating the most current work in the field of resource sharing, we identified that there are limitations and challenges in the available techniques. Most of these technologies depend on centralised points of control or service provides which may constitute a single point of failure. This thesis argues that an open distributed system architecture modelled as a multi-agent system would overcome the centralisation problem. This model provides for node/agent autonomy and decentralised control in addition to the ability for self-adaptation without external interference. However, in open distributed systems, the process of resource discovery is rather challenging hence no central service providers. As the network may scale up largely, the resource search process may take an inordinate amount of time. To address this issue, this thesis introduces distributed resource search algorithms enabling agents to search for resources in a time constrained manner. As such environments keep changing; an agent cannot rely on static information. Hence, this thesis introduces a self-organisation algorithm that the agents can use and adapt their connections to maintain better connectivity to the most valuable connections in terms of resources. Finally, this thesis presents an agent organisation formation and search protocol which facilitates the process of searching for resources across a network of agent organisations instead of searching among individual agents.

During the last few years, systems have emerged to perform large-scale computation and data storage over IP-enabled data communication networks using Grid middleware technology. Grid middleware integrates the computational resources, which may be distributed geographically, over networks. These Grid implementations are currently supported only over IPv4. The next generation Internet Protocol - IPv6 - is replacing IPv4 with a number of improvements. Since IPv6 is expected to become the core protocol for next generation networks, Grid computing systems should be able to continue to work as the lower-layer network protocols migrate to IPv6. Therefore, we studied in depth what needed to be done to integrate IPv6 functionality into Grid middleware here we include both Grid middleware itself and its interface to the underlying networking environment. We have also given consideration to how a Grid implementation can be made to work in heterogeneous IPv4/IPv6 networks. We have used the Globus Toolkit as our working example of a Grid implementation. However, the mechanisms and approaches for integrating IPv6 into the Globus Toolkit are generic. It should cover the integration of IPv6 into most other Grid implementations and even to other IP-based applications. Another aspect of my work relates to the provision of mobility support for Grid middleware, since a lot of Grid resources and users have to be mobile in the wide-area distributed computing environment. Amongst the many mobility solutions, Mobile IP we find the most suitable it has two main advantages in its provision of mobility support in the lower-layer network infrastructure. Firstly, it separates the mobility operations from upper-layer applications, here the Grid middleware. No resultant changes are required in either the applications or the Grid implementations. Secondly, Grid hosts can maintain the identities, so that they can work continuously. The use of Mobile IPv6 rather than Mobile IPv4 is more efficient. This shows that our effort in making Grid middleware IPv6-enabled has brought advantages into the Grid computations. The success of running Grid middleware over Mobile IPv6 builds up only the lower infrastructure for the mobile-enabled Grid by solving the transparent access and handover issues in mobility scenarios. The Grid needs to be modified and improved further in order to work effectively in the mobile environment. The study indicated the major Grid-relevant issue in mobility scenarios is that the status of the Grid changes frequently. Therefore, we introduce a dynamic Grid resource discovery mechanism. Then, we categorise these important characters into four aspects. They are monitored and parameterised dynamically allowing Grid middleware to assign Grid resources dynamically. Eventually, we provide Grid resource mobility functions. Finally, while we have concentrated on the Grid environment, most of methodology and the generic approach apply equally well to other environments.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.