Service Discovery and Semantic Overlay Network Creation in DBGlobe

1
Service Discovery and Semantic Overlay Network
Creation in DBGlobe

• University of Ioannina

4th DBGlobe Meeting Paris, June 23, 2003
2
Outline

• Part 1 Service Discovery
• Part 2 Semantic Overlay Networks
• Part 3 Ontologies

3
Outline

• Part 1 Service Discovery
• Part 2 Semantic Overlay Networks
• Part 3 Ontologies

4
System Architecture

• PMOs are attached to Cell Administration Servers
  (CASs)
• CASs are responsible for the service discovery
  process
• XML data or XML-based service descriptions

5
Service Discovery

• Each CAS maintains data summaries (e.g. Bloom
  Filters) to assist query routing

B
SumB
A
C
SumC
6
Multi-level Bloom Filters

• Hash-based indices that extend Bloom filters to
  support the evaluation of path queries.
• Two approaches Breadth and Depth Bloom Filters
  that rely on different ways of hashing an
  XML-tree.
• Compact structures
• Appearance of false positives

7
Performance Results

• Multi-level Bloom filters outperform in terms of
  false postives Simple Bloom filters in evaluating
  path queries.
• For 2 of the total size of the data, multi-level
  Bloom filters evaluate path queries for a false
  positives ratio below 3
• Breadth Blooms work better than Depth Blooms.
• Depth Blooms require more space but are suitable
  for special type of queries.

8
Filters Distribution

• Peers organized into hierarchies connected
  through a main channel
• Each server maintains
• a local filter
• a merged filter of the filters in its sub-tree.
• If it is a root-peer(connected to the main
  channel) a merged filter for every other
  root-peer

9
Distribution Hierarchical Organization
Node C Local filter Merged filter E? F ? G ?
H Root filters A, B, D
10
Outline

• Part 1 Service Discovery
• Part 2 Semantic Overlay Networks
• Part 3 Ontologies

11
Content-based organization

• Group peers together according to their content
• Use filter and not data similarity for efficiency
• When a peer joins the system
• it broadcasts its local summary and attaches to
  the most similar peer available

12
Bloom Filter Similarity

• Nodes organized according to Bloom Filter
  Similarity
• Measure similarity measure based on the
  Manhattan distance metric.
• Let two filters B and C of size m
• d(B, C) B1 C1 B2 C2 Bm
  Cm.
• similarity(B, C) m d(B, C).

13
Bloom Filter Similarity (contd)
B
1
0
0
1
1
0
0
1
C
0
1
1
0
1
0
0
1
similarity(B, C) 8 - (0 1 1 0 1 0 1
0) 4
For multi-level Bloom filters similarity is
defined as the sum of each pair of corresponding
levels
14
Performance Results

• The content-based organization is much more
  efficient in finding all the results for a query,
  than the proximity organization.
• They both perform similarly in discovering the
  first result.
• The content-based organization outperforms the
  proximity one when the nodes that satisfy a given
  query are limited.

15
Current work

• A peer can belong to more than one hierarchies.
• Self-organization--tuning of the system
  predefined threshold.
• Service Discovery
• Locate the right cluster (hierarchy)
• Find the peer in the hierarchy

16
Outline

• Part 1 Service Discovery
• Part 2 Semantic Overlay Networks
• Part 3 Ontologies

17
Ontologies
Ontologies are hierarchies --gt Thus they can be
summarized by multi-level Blooms
18
Ontologies (contd)

• The main issue
• How to locate the matching hierarchy(cluster)
• Just check every root peer.
• Can we use a global ontology to route us to the
  matching hierarchy more efficiently?

19
Thank you