Efficient Processing of XPath Queries with Structured Overlay Networks

Abstract

Abstract. Non-trivial search predicates beyond mere equality are at the current focus of P2P research. Structured queries, as an important type of non-trivial search, have been studied extensively mainly for unstructured P2P systems so far. As unstructured P2P systems do not use indexing, structured queries are very easy to implement since they can be treated equally to any other type of query. However, this comes at the expense of very high bandwidth consumption and limitations in terms of guarantees and expressiveness that can be provided. Structured P2P systems are an efficient alternative as they typically offer logarithmic search complexity in the number of peers. Though the use of a distributed index (typically a distributed hash table) makes the implementation of structured queries more efficient, it also introduces considerable complexity, and thus only a few approaches exist so far. In this paper we present a first solution for efficiently supporting structured queries, more specifically, XPath queries, in structured P2P systems. For the moment we focus on supporting queries with descendant axes ("//") and wildcards ("*") and do not address joins. The results presented in this paper provide foundational basic functionalities to be used by higher-level query engines for more efficient, complex query support.

An Internet-Scale Service for Publishing and Locating XML Documents

“…Most of the previous work use a form of path summarization (e.g., DataGuide summarization [4], root-to-node paths [3], bloom-filters [11], path hashes [6]) to index XML documents and locate them in a P2P network. Therefore, a complex XPath expression is first decomposed into simple paths and each path is processed separately by issuing separate lookups into the P2P network.…”

“…Recently there has been a rising interest for P2P systems that adopt XML as their data model [3] [4] [5] [6][7] [8]. The ability to model the underlying heterogeneity of data sources and expressiveness of query languages such as XPath and XQuery make XML a suitable choice.…”

Cooperative XPath caching

“…This approach is not appropriate for indexing cached XPath queries, since caches and thus content, changes dynamically. In [17], a distributed index for XPath queries is proposed. The index is based on a distributed binary tree and queries are routed based on their prefixes.…”

“…The main difference with our prefix trie is that [17] is used for indexing the document collections maintained by the peers, while our prefix trie indexes cache content, which significantly changes query processing. Additionally, in [17], P-Grid [1] is used as the DHT, which requires restructuring the overlay to adapt to the indexed dataspace. In a cooperative cache this would occur often, since the cached data changes dynamically according to the query workload.…”

ViP2P: Efficient XML Management in DHT Networks

“…Several DHT-based XML data management platforms [3,9,11] are only concerned with locating in the P2P networks the documents relevant for a query. All the peers which may hold results then locally evaluate the query, leading to high query traffic and peer overload.…”