Network Flow Bandwidth Requirements FLOW AVG.RATE SRC. TYPE Video 4Mbps UCL Conference transmission DST. DUR. Access Grid Nodes 4-5 hours FREQUENCY NUM. START YEAR 1 per day 2002002 250/year UK GRID Networking Research Projects GRID Infrastructure Service Infrastructure Network Infrastructure GRS, GRID resource management ‘ FutureGRID, P2P architecture GridMcast, Multicast-enabled data distribution MB-NG, QoS Features GRIDprobe, backbone passive monitoring at 10Gbps MB-NG (Managed Bandwidth Next Generation) Objectives: To demonstrate managed bandwidth and QoS services in a Grid environment (over MPLS); To demonstrate sustained gigabit/s scale data transport across the wide area network (over MPLS); To engage in tests with collaborative partners in the EU and USA. Partners: UCL, Man, CLRC, UKERNA, Cisco GridProbe Objectives: Development of an open and extensible architecture for network monitoring based on modularity. Performance requirements are for monitoring 10 Gbps links. Development of novel high-performance measurement tools. Which will allow any desired analysis to be performed upon the results Extrapolation and analysis of architectural requirements, constraints, and trade-offs which will be necessary to monitor networks of 40 Gbps and beyond, and develop a strategy to achieve this goal. Partners: Cambridge, Intel, Waikato DAG, Broadcom, Xilinx, NTL, Thus GridMcast Objectives: Investigation of the implementation issues of multicast transfers within the Grid framework. Provide a far superior method of software distribution, data replication and mirroring, and access to shared data for visualisation and analysis. All of these tasks are key to the success of PPARC's major Grid- based projects, GridPP and Astrogrid, which foresee Grids of international scale, with thousands of nodes. Partners: Cambridge e-Science Centre, GridPP(EU), ATLAS (CERN) FutureGRID Objectives: Deploy Xenoservers Active nodes at Access Grid sites (all e-Science Centres) Port the Pastry message routing and the Scribe event notification frameworks to Xenoserver with accounting and resource management of network usage (multicast congestion control) Design suitable API based on GRID GGF models WSDL/OGSA Refinement and implementation of application set (vic, rat, fcast, powerpoint) to new infrastructure, including gateways to legacy “native” multicast versions. Partners: Cambridge (CLab, Microsoft Research). GRID Resource Scheduling (GRS) Objectives QoS and network features integration in the GRID resource model. Dynamic Requests of Bandwidth for capacity from end-users. Micro-management of QoS at edge sites. Highly decentralised, multi-domain QoS management under local administrative controls Partners UCL, Cambridge, DataGRID