Abstract (EN)

The paper presents a complete framework for spatial indexing support in a distributed setting. We consider a shared-nothing environment where a set of servers provides independent storage and computational services. Servers only communicate through point-to-point messaging, and constitute a non-structured network (i.e., non-central server or "super peer"). These features cover two popular architectures, namely a strongly connected cluster of servers, and P2P networks. Our proposal extends the recently proposed "Scalable Distributed Rtree (SD-Rtree)" structure with new algorithms and protocols. More specifically, we introduce a nearest-neighbors algorithm, a load balancing method and a termination protocol. The result constitutes a set of functionalities for distributed spatial indexing that matches those commonly found in centralized architectures.