Load balancing the GRID ad-hoc routing protocol

Abstract

Grid is a mobile ad hoc routing system with significantly better scaling properties than previously designed protocols in networks with a uniform distribution of nodes. It achieves scalability by building a distributed location database in which memory requirements are apportioned fairly to all nodes in the network. In networks of spatially non-uniform node distribution, however, a small fraction of the nodes need to store a disproportionate amount of location information, posing a significant problem for nodes with limited memory such as small handheld devices. We propose a new location service to improve the scalability of Grid while preserving its fundamental design. Our new service, GLS2, ensures protocol correctness despite limited memory constraints. To preserve high success rate and improve query path efficiency, GLS2 applies a new location query algorithm which allows nodes to drop location information if necessary. Simulated tests demonstrate that GLS2's efficiency and correctness are preserved in situations of limited memory as well as those of extremely uneven node distributions while still achieving proper load-balancing.