CISS: An Efficient Object Clustering Framework for DHT-Based Peer-to-Peer Applications

VBI-Tree: A Peer-to-Peer Framework for Supporting Multi-Dimensional Indexing Schemes

“…Being a structure that has some similarity with the kd-tree [3] and grid-file [9], CAN can be used to directly index multi-dimensional data in its natural space. Most other systems such as [16,12] use space filling curves to map multi-dimensional data to one dimensional data. After that, an overlay network is used to index that one dimensional data.…”

“…Multi-dimensional indexes such as the R-tree [8] and R * -tree [2], and high-dimensional indexes such as the M-tree [7] have been well tested and are widely accepted as robust indexes in centralized systems. Even for P2P systems, there have been a few proposals to support multidimensional indexing [18,12,21,19]. Most systems sup- * Supported in part by NSF grant EIA-0303587 † Supported in part by IDA CCC grant as part of BestPeer [13] project porting multi-dimensional data indexing in centralized database are based on tree structures, which have many robust properties such as concurrency, scalability, and adaptivity.…”

Replica-aware, multi-dimensional range queries in Distributed Hash Tables

“…To avoid load imbalance, Squid relies on the fact that the d-dimensional keyword space is sparse and so the data items are assigned to peers roughly in the same way. Multidimensional indexing via a Hilbert SFC is also employed by the CISS framework [29]. Assuming a hierarchy for each attribute, performing a range query requires the node responsible for the first key of a candidate cluster to be looked up, before forwarding to succeeding peers, until all relevant objects are retrieved.…”

Multi-Attribute Range Queries on Read-Only DHT

“…In d-to-one mapping scheme such as Squid [16], CONE [15], ZNet [17], SCRAP [14] and CISS [18], we consider a d-attribute resource as a coordinate point in a d-dimensional attribute space. The point is mapped to a key in a onedimensional identifier space using a d-to-one mapping function such as space-filling curve (SFC).…”

“…An example of 2-attribute range queries is to find a compute resource whose cpu = P 3 and 1 GB ≤ memory ≤ 2 GB. However, current approaches such as distributed inverted index [8]- [12], d-to-d mapping scheme [13]- [15], and d-to-one mapping scheme [14]- [18], assume that the underlying DHT distributes key-value pairs across the overlay network.…”