This paper proposes a distributed peer-to-peer data lookup technique on DHTs in order to serve range queries over multiple attributes. The scheme, MARQUES, uses space filling curves to map multi-attribute data points to a one-dimensional key space and thus effectively converts multi-attribute range queries into a consecutive series of one-dimensional keys. These keys are then used to place or lookup data objects over a DHT.
Space filling curves preserve locality of attribute values and thus helps greatly in facilitating range queries in terms of the number of nodes to be searched to serve a given range query. MARQUES, although can be instrumented with any space filling curve, has been implemented with two curves, namely Z-order curve and Hilbert curve, and uses a multi-level variant of Chord, a popular DHT, as its underlying overlay. Simulation results on OMNET++ show that MARQUES successfully answers range queries with significant efficiency in terms of message overhead and query latency.