A better way to negotiate for testbed resources

Resource allocation is an increasing challenge for distributed network testbeds as computational and network resources are involved. Testbed designers have moved to a query-based model: clients provide a declarative description of their desired resources, and the provider allocate specific resources to meet the request. In this paper, we describe an new approach to negotiate testbed resources between clients and testbed providers: the clients specify their requests as constraints, and the providers reply with resource allocations expressed also as declarative set of constraints on resources. This gives providers more flexibility in late-binding of resources to requests, and opens up a wide design space to optimize resource allocation for efficiency, cost, utilization, or other metrics. Our simple first experiments suggest that the late-binding of resources enabled by representing resource reservation as constraints achieves better network resource utilization compared to the fixed assignment solution.