Back to search

National Technical University of Athens (NTUA)

Προσαρμοστική διαχείριση και αναζήτηση δεδομένων ευρείας κλίμακας σε κατανεμημένα συστήματα

Abstract

dc:description

The tremendous increase of managed data by a variety of applications is a new trend observed more and more in our digital era. The ability to handle and analyse large amounts of data efficiently is a requirement posed strongly by scientific and business disciplines and the Web community. This trend leads to the adoption of distributed solutions for data management aiming at building scalable and fault-tolerant systems combining the power of multiple autonomous resources. Peer-to-Peer networks greatly contribute to the design of decentralized systems capable of dynamically adjusting to changes of their topology. A major class of existing Peer-to-Peer networks is the one referring to structured overlays that implement a Distributed Hash Table (DHTs). The efficient lookup functionality provided by the Distributed Hash Tables has made them popular among Internet-scale applications for content publishing and sharing. The main goal in this dissertation is the development of data management techniques for large-scale data hosted by scattered resources. In this context, novel methodologies for efficient organization, indexing, searching and updating of data are introduced. A common property of the explored data is the use of concept hierarchies, which offer the capability for organizing it at different levels of abstraction. The exploitation of concept hierarchies greatly helps in the organization and reuse of information and contributes to more effective processing of aggregate queries. The proposed techniques enable the organization of such data in a manner that preserves the semantics of the hierarchies, while they assign it among the nodes of a DHT substrate. The applied method for query processing utilizes a distributed indexing scheme allowing peers to dynamically detect the prevailing trends in incoming queries and adapt the indexing granularity. Re-indexing operations towards more general or more detailed levels can be performed individually on a per node basis for each stored hierarchy to improve the performance of the processing and expedite the retrieval of results on variable aggregation levels. Another important aspect addressed in this approach is the support for online updates of stored items and insertions of new ones without impeding the operation of the system. Apart from the evaluation of the proposed techniques with synthetic query workloads following uniform and skewed distributions, this approach is also motivated by the use case of a Grid Information System. A fully decentralized scheme is developed that creates, queries and updates large volumes of hierarchical data on-line and can be considered as a viable solution compared to the traditional information systems comprising of centralized and hierarchical structures. The derived methods are further extended and enhanced to enable the management of data annotated by concept hierarchies in multiple dimensions. The outcome is the implementation of all required mechanisms for a fully functional system handling data following the specific structure. A significant asset of the system differentiating it from existing approaches is the restriction for a global, rigid schema followed by all data is eliminated. The proposed scheme manages partially structured information and search strategies are described, which mainly focus on the resolution of aggregate operations over multiple dimensions. Apart from the adoption of the highly adaptive mechanisms performing re-indexing operations, a technique for the pre-computation of combinations of stored values is analysed for partial materialization of views according to the incoming queries. The introduced techniques for the management of semi-structured, multi-attribute data are also applied in the design of a system for semantic search and retrieval. The paradigm of Linked Data, which is widely used for publishing large datasets from different resources on the Web is studied. The resultant system is a distributed platform that serves the needs for integrating, indexing and querying data published in the form of Linked Data.

Degree

thesis:*
Grantor dc:publisher
National Technical University of Athens (NTUA)
Year dc:date
2012

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Asiki, Athanasia
  • Ασίκη, Αθανασία

Subjects

dc:subject × 16

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/27477

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Asiki, Athanasia; Ασίκη, Αθανασία. Προσαρμοστική διαχείριση και αναζήτηση δεδομένων ευρείας κλίμακας σε κατανεμημένα συστήματα. National Technical University of Athens (NTUA), 2012. http://hdl.handle.net/10442/hedi/27477