Addressing Scalability in a Laboratory-Based Multihop Wireless Testbed

The MeshTest wireless testbed allows users to conduct repeatable mobile experiments with real radio hardware under controlled conditions. The testbed uses shielded enclosures and a matrix switch of programmable attenuators to produce multi-hop scenarios and simulate the effects of mobility and fading. Previous work focused on the theory and performance of a single-switch testbed. Connecting more than 16 nodes requires multiple matrix switches and introduces theoretical and practical challenges. In this paper we examine, in theoretical and practical terms, two potential designs for a scalable version of the MeshTest testbed, and identify one design that seems to provide the most promising test environment. For this type of multi-switch system, performance improves a great deal if nodes that are close together in the physical scenario are also connected to the same RF switch. Rather than restrict the mobility of the nodes, we show how software virtualization can be used to migrate running node images from one switch to another to maintain the proximity of the nodes attached to each RF switch and improve testbed performance. We also describe our new testbed control architecture, that is an important part of making MeshTest scalable.