Link Scheduling for Scalable Data Aggregation

We explore the link scheduling optimization problem in the context of scalable in-network data aggregation, extending results for broadcast networks to routing in general networks. The primary vehicle for resource preservation is transmission suppression. For certain types of queries, nodes can avoid transmitting records if they can locally infer that their data is not needed to execute the query. We introduce a novel protocol paradigm for duplicate-insensitive exemplary monotonic (e.g. MIN and MAX) data aggregation queries. Performance of query execution in these networks is measured through collective expected number of transmissions in the network, and is linked to the minimum and maximum connectivity (delta, Delta) of nodes in general graphs. Nodes can reduce transmissions and save power by forcibly broadcasting partial results to the network during data collection. An algorithm running in time O(nm) is presented which achieves O((n In n In Delta)/delta) expected transmissions.