diane project michael klein, birgitta könig-ries multi-layer clusters in ad-hoc networks - an...

25
DIANE Project Michael Klein, Birgitta König-Ries http://www.ipd.uni-karlsruhe.de/DIANE Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service Discovery Universität Karlsruhe Institute for Program Structures und Data Organization Universität Karlsruhe GERMANY International Workshop on Peer-to-Peer Computing co-located with the NETWORKING 2002 Conference May 24 th , 2002 – Pisa, Italy 1/15

Post on 18-Dec-2015

214 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

DIANE Project

Michael Klein, Birgitta König-Ries

http://www.ipd.uni-karlsruhe.de/DIANE

Multi-Layer Clusters in Ad-hoc Networks -

An Approach to Service Discovery

Universität KarlsruheInstitute for Program Structures und Data Organization

Universität KarlsruheGERMANY

International Workshop on Peer-to-Peer Computingco-located with the NETWORKING 2002 Conference

May 24th, 2002 – Pisa, Italy

1/15

Page 2: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

Our Scenario

Anna

More on SQL?

Official SQL Slides1 - 2 - 4

Summary on 2PC

Exercise Sheet on UML

Exercise Sheet on SQL

Solution to SQL Sheet

2/15

Page 3: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

Problems with mobile Ad-hoc Networks

• Highly dynamic topology due to • node movement • node fluctuation• appearing obstacles Routing difficult

• No dedicated server, no physical infrastructure No central service directory

3/15

Page 4: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

How to search for services?

Product search in a shopping centre

• Similar products are fixedly placed in physical proximity• Search by exploring the places around a similar product

?

Product search in an ad-hoc network• No explicit corelation between semantical and

physical proximity• Temporal changes in service offers and location

4/15

Page 5: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

Our Approach: Multi-Layer Clusters

Idea• Build clusters of devices that locally combine

semantical and physical proximity• Build supercluster of clusters by relaxing

proximity demands

5/15

Page 6: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

Semantical Proximity by an Ontology (1)

• Use a common ontology as a measure for proximity

• Use only isSubTopicOf and isDescribedBy relations

• Assumption: Each device offers one document, which can be described by one leaf term of the ontology

database

object oriented model

relational model

isSubTopicOf

rel. algebra

SQL OQL

isSubTopicOf

isDescribedBy

Two services/clusters are semantically similar iff. they belong to the same ontological term

6/15

Page 7: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

Physical Proximity by Radio Reachability

Device a reaches Device b iff.a is currently able to send data to b directly

a

b

Cluster A reaches Cluster B iff.there is a member m1 in A and a member m2 in B such that m1 reaches m2 ( gateway nodes)

A B

m1m2

7/15

Page 8: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

Clustering (1)

Step 1Form a layer 1 cluster from devices thata) are semantically similar

(= are described by the same ontological term)

b) and are physically close(= form a connected reachability graph)

select.doc

sql1.ppt

sql3.ppt

projection.pdf

selection.pdf

division.doc

relAlgebra1.ppt

sql2.ppt

insert.doc

update.doc

8/15

Page 9: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

Clustering (2)

Step iForm a layer i cluster from layer (i-1) clusters thata) are semantically similar

(= share the same supertopic term in the ontology)

b) are physically close(= form a connected reachability graph)

SQL

SQL

Rel. Algebra

Relational Model

9/15

Page 10: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

Service Discovery

The goal is to have a function

Device findService(Service s)

which• searches for a Device offering Service s• can be called from an arbitrary device in the network• can be used to find an arbitrary Service s• can be implemented locally (not centrally)

But we have:• Very basic functions on devices:

• 1. check if service request s matches• 2. send message to a reachable device

• Clustering of the devices

IdeaLayer Architecture: Break down the complex functionality in several steps.

User view

System view

gap

10/15

Page 11: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

Layer Architecture

Device Layer 0

Cluster Layer 1

Cluster Layer 2

Root Layer n

Cluster Layer (n-1)

ViewSearch function

Small Clusters ofterms of Level 1

Single devices (only on the current device)

(only in the current cluster)

(only in the current cluster)

(only in the current cluster)

Device findService(Service s)

(everywhere)

Clusters ofterms of Level 2

Big Clusters ofterms of Level n-1

One cluster ofthe root term

Send function

(only to reachable clusters)

sendTo(Node n, Message m)

--

(only to reachable clusters)

(only to reachable clusters)

(only to reachable devices)

given

11/15

Page 12: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

ExampleThe Ontology

?

findService( )

12/15

Page 13: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

ExampleExampleThe Ontology

findService( )

findService( )

12/15

Page 14: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

ExampleThe Ontology

Different routing methods:• Flooding

• Cycling (Ring)

• Direct (Table)

findService( )

12/15

Page 15: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

ExampleExampleThe Ontology

sendMessage( )

findService( )

12/15

Page 16: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

ExampleThe Ontology

sendMessage( )

findService( )

12/15

Page 17: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

ExampleThe Ontology

findService( )

sendMessage( )

findService( )

12/15

Page 18: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

ExampleThe Ontology

sendMessage( )

sendMessage( )

findService( )

findService( )

findService( )

12/15

Page 19: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

ExampleThe Ontology

sendMessage( )

findService( )

12/15

Page 20: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

ExampleThe Ontology

sendMessage( )

findService( )

findService( )

12/15

Page 21: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

ExampleThe Ontology

12/15

Page 22: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

ExampleThe Ontology

12/15

Page 23: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

Advantages of the Approach

• Naturalness only semantical and phyisical proximity, no parameters

• Decentralization no central device

• Resource-Awareness searches local clusters before accessing distant ones

• Adaptability to local network stability dynamically adapts exploration strategy

• Fault Tolerance by changing exploration strategy

13/15

Page 24: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

Future Work

Some open questions:

• Management of administrative data(routing tables, ring predecessors and successors, border nodes, service descriptions etc.)

elect cluster head in each cluster replicate on all cluster members (lazy replication)

• Performance Implementation in simulator QualNet

14/15

Page 25: DIANE Project Michael Klein, Birgitta König-Ries  Multi-Layer Clusters in Ad-hoc Networks - An Approach to Service

Thank you!

More information on our project web page:http://www.ipd.uni-karlsruhe.de/DIANE/en

Are there any questions?

Thank you for your attention!

15/15