Network QoS provision for distributed grid applications

Grid computing provides a global-scale distributed computing infrastructure for executing scientific and business applications. An important requirement is the need to make this infrastructure appear as a single logical coordinated resource – which in reality consists of a variety of resources aggregated across different administrative domains. The aggregation of network resources is often undertaken over a ‘best effort’ infrastructure as provided by the Internet – however many applications, which necessitate soft-real time constraints, such as collaborative working or remote visualisation, require more stringent traffic guarantees. Managing Quality of Service (QoS) requirements across these aggregated resources therefore becomes an important concern. Such QoS criteria must extend to computational, data and network resources, and are often expressed in a Service Level Agreement, multiples of which may co-exist over the entire collection of resources. In this paper we focus on network QoS as part of our wider work on the Grid QoS Management (G-QoSm) framework. Provisioning of such QoS support is provided via the Differentiated Services architecture, and relies on the use of a Bandwidth Broker (BB) component. Performance results of using the BB alongside other elements in G-QoSm are presented