The Grid Computing and Distributed Systems (GRIDS) Laboratory, University of Melbourne

Over the last few years, several Grids have been set up to share resources such as computers, data, and instruments to enable collaborative research. Various national programs have initiated e-Science projects to enable resource sharing and collaboration among scientists. Such endeavors generally follow a restricted organizational model based on the idea of Virtual Organizations (VOs). These models and dispersed Grid initiatives have resulted in islands of Grids without resource sharing between them. There has been an increasing interest in providing interoperability and interlinking current Grid infrastructures. Nevertheless, much beyond the need for interoperability at middleware level, interlinking of Grids requires peering arrangements, advanced and automated mechanisms for inter-Grid resource allocation, reservation, accounting, and scheduling. In addition, Grids need to adopt mechanisms that enable administrative separation, by allowing networks of networks, similar to many network-based systems such as the Internet, the Web, and numerous social and biological systems.

The current structure of Grids does not follow principles such as the peering between Internet Service Providers (ISPs) present in the Internet. For example, a set of common communication protocols underlies the Internet, but when ISPs peer with one another they consider their policies, economic issues, social and economic impact of peering. Therefore, likewise the peering of ISPs, we need to investigate policies for interlinking of Grids and how the peering arrangements between Grids will be made. In this project we identify the key issues of current Grid technologies that do not allow them to evolve to such an InterGrid level and work to answer some questions such as:

• What kind of structure should the InterGrid have to promote internetworking of Grids?
• What kind of ecosystem, peering arrangements and policies do we need to have?
• What are the coordination mechanisms that we need to put in place to enable the InterGrid?
• What are the incentives for end users, laboratories, organizations, service providers, and Grid facilities in general to engage in such a network of Grids?
Should we consider macroeconomics-based, bilateral, and multilateral agreements when linking such Grids?

We are currently investigating mechanisms and policies for resource allocation across Grids based on InterGrid Gateways (IGGs). The scenario we consider is described in the figure. The peering arrangements between Grids will enable the deployment of Distributed Virtual Environments and Virtual Organisations (VOs) spanning multiple Grids. We hope that the Grid research community will provide us with comments and help us in realising this vision.