a collaborative and semantic data management framework for ubiquitous computing environment...
TRANSCRIPT
A Collaborative and Semantic Data Management A Collaborative and Semantic Data Management Framework for Ubiquitous Computing Framework for Ubiquitous Computing
EnvironmentEnvironment
International Conference of Embedded and Ubiquitous Computing (2004)
Presented By Weisong Chen, Cho-Li Wang, and Francis C.M. Lau
Department of Computer Science, The University of Hong Kong
Summerized By Jaeseok Myung
Copyright 2008 by CEBT
IntroductionIntroduction
Characteristics on Ubiquitous Computing
Distribution
Heterogeneity
Mobility
Autonomy
These characteristics introduce tremendous data management challenges, which cannot be easily overcome by existing solution
Center for E-Business Technology
Copyright 2008 by CEBT
Key IdeaKey Idea
A Guiding Principle behind System Design
Encourage contributions from devices owned by different users
Assumptions
People joining the environment are expected to agree to share their devices
Core Techniques
Ontology-based Metadata
– An effective approach to deal with data diversity in the ubiquitous environment
Incentive-based Routing Protocol
– Provide incentives for devices to contribute to others’ information accesses
– The more contribution a device makes, the more knowledge it will gain
Cooperative Caching
– Maintain local cached copies of the downloaded data and share them with others
– Popular data will be widely cached and unused data will fade away eventually
Center for E-Business Technology
Copyright 2008 by CEBT
Incentive-based Routing ProtocolIncentive-based Routing Protocol
When forwarding queries, nodes record the nodes that initiated the queries
Enhancing the ability of these nodes to serve future queries
When passing the query results to the initiating nodes, the nodes record the nodes providing the results
Center for E-Business Technology
N1 N3N2
Q
M
Q, N1 Q
M, N3M, N3
Copyright 2008 by CEBT
Ontology & MetadataOntology & Metadata
Center for E-Business Technology
Copyright 2008 by CEBT
OntologyOntology
Ontology, O = { C, P, HC, R}
Concepts (C) : Well-defined terms referring to classes(or types) of objects in a particular domain
Relations (P) : Properties of concepts defining the concept semantics
Concept Hierarchy (HC) : A hierarchy of concepts that are linked together through relations of specialization and generalization
R : A function that relates two concepts non-taxonomically, using the relations in P. R(P) = (C1, C2) is usually written as P(C1, C2)
Center for E-Business Technology
Copyright 2008 by CEBT
MetadataMetadata
Metadata, M = { O, I, C, PI, IC, IR }
O : a referenced ontology
I : a set of concept instances
C : a set of concepts (a subset of the concepts in the ontology)
PI : a set of relation instances
IC : I -> C, a function that relates instances to the corresponding concepts
IR : PI -> I x I, a function to relate instances using relation instances; IR(PI) = (I1, I2)
For each piece of metadata, there’s one concept instance that serves as the identifier of the described data
MI : Central Concept Instance
MC : Central Concept
The query structure and the meaning of each element are same as those of the metadata
The query allows wildcard instance (denoted as I*)
Center for E-Business Technology
Copyright 2008 by CEBT
Query ProcessingQuery Processing
Center for E-Business Technology
N1
MC
MC
MC
M M M
M
MM
Q
Msim(Q, M)
Copyright 2008 by CEBT
Metadata Similarity (1)Metadata Similarity (1)
The degree that metadata M2 is similar to M1 is given by the
following formula, where IM2 denotes the concept instance set
of M2, excluding the central concept instance M2I
The similarity level between two concept instances is given by the following formula, where INIL means that the concept
instance does not exist
Center for E-Business Technology
Copyright 2008 by CEBT
Metadata Similarity (2)Metadata Similarity (2)
Similarity between two concepts in a concept hierarchy
T. Andreasen et al., From Ontology over Similarity to Query Evaluation, 2003
Center for E-Business Technology
SC(Publication) = {Publication, Report, Book}
SC(Report) = {Publication, Report}
Copyright 2008 by CEBT
Performance EvaluationPerformance Evaluation
Parameter Settings
Center for E-Business Technology
Copyright 2008 by CEBT
Ontology vs. Keyword SearchingOntology vs. Keyword Searching
In both cases, as more queries are issued, the cached data contribute more to the overall hit ratio
Ontology-based searching has far superior performance
Center for E-Business Technology
Copyright 2008 by CEBT
Effect of Cache Replacement and Query Effect of Cache Replacement and Query PatternsPatterns
Random : no predefined pattern
Interest-based : only for some limited number of concepts
Popularity-based : generate queries according to what are popular
Center for E-Business Technology
Copyright 2008 by CEBT
Comparison with Other SystemsComparison with Other Systems
Proposed system and FreeNet have much better performance than others
FreeNet only supports exact ID matching
Center for E-Business Technology
Copyright 2008 by CEBT
Conclusion and Future WorkConclusion and Future Work
Characteristics on Ubiquitous Computing
Distribution
Heterogeneity
Mobility
Autonomy
A Collaborative and Semantic Data Management Framework for Ubiquitous Computing Environment
In this paper, the authors have assumed that complete ontology knowledge is available at each device, which is not always possible in the ubiquitous computing environment
Center for E-Business Technology
Copyright 2008 by CEBT
DiscussionDiscussion
Comparing with P2P Architecture
Is the incentive really attractive?
Hit Ratio is OK, but the propagation cost must be expensive
Center for E-Business Technology