1

1
Data Management in MANET

Dr. Noman Islam

https://sites.google.com/a/nu.edu.pk/noman-islam/
http://facebook.com/sir.noman.islam
 Introduction
2

We are living in the era of computing where data has

become a critical and most valuable asset
The progression of technology and the emergence of
mobile and ad hoc networking have further enabled
the creation and consumption of information on move
Mobile Ad Network (MANET) is defined as a network
erected among a number of self organized mobile
nodes, on the fly
These nodes interact and exchange data with each
other in P2P style to solve complex user problems
 Introduction contd
3

Several useful applications are thus possible

However, to truly realize these applications requires

the need for addressing an array of new issues and
challenges related to management of data
 Mobile, resource
limited, unreliable and
heterogeneous nodes

Low capacity,
asymmetric and
instable links

Dynamic,
infrastructure-less and
data intensive
environment

4 Mobile Ad hoc Network Vs Conventional Networks

 Introduction contd
5

Unlike conventional client-server computing, the

P2P communication generates massive volume of
data
Despite the immense need, there has been
shortage of a comprehensive data management
framework for MANET
 Introduction contd
6

Hence, this work presents ALADIN. The framework

utilizes:
Inter-layer communication across layers of protocol stack
to curtail network dynamism
i.e. Coordination, interaction and joint operation of protocols
across different layers
Fractional views of semantic information to ensure
scalability
Maintaining ontology documents as needed
Correlation among data items to optimize various data
management operations
There are inherent associations among data items that can
be computed based on usage log
 Research Objectives
7

To propose a comprehensive framework for MANET that

addresses key issues of data management by utilizing:
1. Cross-layer communication
2. Multi-level ontologies
3. Inherent relationship among data

This work doesnt provide:

Detailed discussions and evaluations of other issues of data
management (replication, security management etc.). But it
does sketch solutions for these issues
 Problem Area
8
9 Background and Literature Review
1. Introduction to Data Management
2. Data Management Issues
3. Data Management Frameworks
 What is Data Management?
10

We can define data management as:

a set of actions performed on the data considering the end
users preferences and requirements
Based on the literature review, we can identify following
major steps/issues in data management (Perich, Joshi et al.
2006):
Data representation
Data discovery
Cache Management
Query Processing and Optimization
Data Replication
Naming
Transaction Management
 Data Management Issues
11

Data Representation - the mechanism to describe

the data such that it can be identified, discovered
and accessed by any local or remote consumer
Challenges
Heterogeneity
Proprietary data formats, Semantic differences
Absence of universal schema
Unbounded/open nature, Resource limitations, Size of schema
Solution
On-demand discovery and maintenance of fractional
schema
 Data Management Issues (contd)
12

Data Discovery - the actions performed for locating the

owner of data considering the preferences specified by
the consumer
Challenges
Lack of dedicated directory servers
Spatiotemporal variations
Frequent disconnections
Solution
Directory-less approaches
Cross-layer operation
Exploit data correlations to store potentially useful data in
advance
 Data Management Issues (contd)
13

Cache Management To locally save a data item for

future use after it has been discovered
Challenges
Lack of prior information about nature of data. Locality of
reference proposed for conventional systems will not work
Frequent disconnections
Frequent state changes
Solution
Exploit data correlations to compute relationships among data
items
Cross-layer operations i.e. utilizing network events for managing
consistency of data
A summary of existing solutions for data management issues
 Data Management frameworks for
16
Mobile Ad hoc Network
Alfanao et al. (2012) a method for managing operations and data in
MANET
CDMAN (Martin and Demeure 2008) a framework for improved data
sharing by structured, segmented data
DRIVE (Zhong, Xu et al. 2008) a system for opportunistic information
sharing in vehicular networks
Ad-hoc InfoWare (Sanderson, Skjelsvik et al. 2007) a framework for
knowledge management in rescue scenarios
AmbientDB (Fontijn and Boncz 2004) a relational database abstraction
over ad hoc networks of heterogeneous peers
MoGATU (Perich, Joshi et al. 2006) a P2P data management framework
based on cross-layer operations, context management etc.
CHaMeLeoN (Shahram Ghandeharizadeh 2006) a framework to
facilitate exchange of text and continuous media over ad hoc networks
Summary of Data Management frameworks for MANET
 Gap Analysis
18

1. There has been shortage of a comprehensive data

management framework
2. Most of the solutions ruminated on a subset of issues
of data management. Key issues of data management
are:
Data Representation
Data Discovery
Cache management
3. None of the solutions exploit correlation and multi-
level ontology for data management
4. Only few solutions are based on cross-layer
communication
19

19 Proposed Framework
1. Introduction
2. Design Principles
3. Block Diagram
4. Description
 ALADIN
20

In this work an Association Rules Based Network

Layer Data and Information Management System
(ALADIN) for MANET has been proposed
It comprises of components for:
1. Network layer, correlation based data discovery

2. Multi-level, ontology based data

representation
3. Correlation-based cache replacement and
cross-layer consistency management
 Frameworks Design Principles
21

ALADIN rests on following three points:

1. Exploits Correlation Information
There are some inherent associations/correlation among data
items, as has been verified (Appendix 1)
Two data items are correlated if they have been used together for
reasonable number of times. Correlated data items are anticipated
to be used together in future
Correlation can be exploited for:
Associative advertisements proactively advertise data to neighbors.
Neighbors will store these advertisements
Piggybacking of information piggyback correlated information with
current data request
Priority based cache replacement keep those data items in cache
that are correlated with current request
 Frameworks Design Principles
22
(contd)
2. Exploits cross-layer operations
Network layer data discovery data items and routes are
discovered simultaneously
Cache Consistency Management exploiting network
layer events for reactively manage data states
3. Multi-level data representation
On-demand discovery and maintenance of semantic
information use two level ontology. Keep the core
ontology at nodes. Extended ontologies can be
discovered when needed
Proposed Data Management Framework
 Description
24

Logging Component
Maintains the fractional history of data requests in the
form of sessions
Association Rules Mining Component
Calculates the correlation among data items by running FP-
Growth algorithm (Han, J et al. 2004) on the log database
FP-Growth is a tree-based data mining algorithm to
compute frequently used itemsets
The output of FP-Growth algorithm is the set of frequent
itemsets representing what data requests have been
mostly consumed together in past
 Data Discovery Component
25

Discovery component is responsible for discovery of data

items from other nodes on the network
It comprises of two subcomponents:
A proactive discovery component that periodically advertises its
data and route via SADV messages
A reactive discovery component to discover the desired
information based on the delivery mechanism of any reactive
routing protocols
The proposed discovery component discovers a data item
and corresponding routes to provider simultaneously
This leads to reduced latency of the overall discovery
process
 Proactive Component
26

Periodically advertises data items at regular intervals to neighbors

Listener component at neighbors can accumulate the received
advertisements
Advertisement (SADV) includes details about the data items along
with corresponding routes to access these data items
The advertiser can operate in two modes:
Correlated Mode
Advertise the catalog entries that are correlated to each other
Correlation is determined based on the itemsets generated by the mining
component
Non-Correlated Mode
Advertise the catalog entries in round robin style
Proactive Discovery Component
 Reactive Component
28

Reactive component performs on-demand discovery of any

requested data item
ALADIN presents two discovery algorithms based on following
reactive routing algorithms:
AODV based Discovery
DSR based Discovery
Source node prepares a discovery request (SREQ) and sends to
adjacent nodes
Intermediate nodes propagates the request until it is answered by
the destination node
In case of AODV, a temporary reverse route is used
In case of DSR, the traversed hop is appended with the original request
A SREP is then generated. SREP is sent back through the same
route used for propagation of SREQ
Discovery Request

Discovery Response

AODV based Reactive Discovery

 Using Correlation Patterns to Improve
30
Discovery Process
The correlation computed by mining component can be
exploited to prophesize about likely future requests and
answers of these guessed requests can be piggybacked/
attached with current request
While generating a discovery response, the replying node
appends the details of data items related to request with
current response
31 Using correlation patterns for piggybacking
 Data Representation
32

MANET suffers from issues of heterogeneity due to its open

nature
Ontology can be used by nodes to reach to a common
agreement of the domain terminologies
However, the use of a global ontology is not viable due to:
Limited resources
Nodes of varied origins
 Data Representation (contd)
33

ALADIN recommends
maintaining two-level
ontology:
Core Ontology
Available at every node
Ext Ontology
Fetched as needed from the
network by invoking the discovery
component
ALADIN Multi-level ontology management approach
A General Purpose Software
Ontology for Ad hoc and
Vehicular Applications has
been proposed
Multilevel approach to maintain ontology
 Core ontology contains
generalized concepts for any
ad hoc and vehicular
applications
Ext ontology extends Core
Ontology and contains
concepts for specialized
applications

An example of Multilevel Ontology

 Cache Management
36

It contains sub-components for:

Cache replacement
Exploitsthe correlation patterns generated by the mining
component to place a newly arrived data item in to the
cache
Victim slot is thus picked from those data items that are
not related to current request
Consistency management
Exploitsthe routing layer to determine the topological
changes in the network.
These topological changes are used to maintain
consistency of cached data
 Cache Replacement
37

C = {c1, c2, c3 } denotes the cache, denotes the

correlation function. Cache Items related to d i.e. Rd
can be expressed as:

Residual Set Sd i.e. data items not related to d can be

written as:

The victim slot is thus picked from those data items

that are not related to d i.e.
38 Proposed cache replacement algorithm
 Consistency Manager
39

The robustness of network layer can be used to ensure

the consistency of items maintained in data catalog
ALADIN proposes the consistency management based
on the existing routing protocols (AODV and DSR etc.)
In case of a link failure, the node (who senses the
failure) attempts to perform local route maintenance
The upstream nodes are then notified depending upon
the status of local maintenance. A failure in local repair
leads to invalidation of routes and data table
The upstream nodes notify their upstream nodes
42

42 Implementation Details
1. Experimental Setup
2. Data sets
3. Demonstration
 Network Simulation
43

All the experiments were conducted on JIST/SWANS Simulator

The ontology has been developed in OWL language using Protg
JENA Toolkit has been employed for ontology processing tasks
 Mobility Models
44

To model the movement patterns of mobile nodes during simulation, mobility models are used

P5 P2 A node chooses a random point

The node moves towards the point with a
P4 uniform but random speed
Pauses for a constant time
P3 Repeats the whole process again
P1

Random Way Point

Node moves according to streets

Streets are defined by real maps of US cities
Mobility is also restricted based on vehicular
congestion and other traffic control mechanisms

Street Random Waypoint Mobility (Straw)

 Datasets
45

Three types of datasets have been used for generating

sessions.
Random Dataset. Picks a data item randomly and
floats on the network
Generating Correlated Data Sessions
The YouTube Dataset: A dataset generated based on
YouTube video requests by users in a campus (Zink, M. et
al. 2008)
The Correlated Data Generator(CDG) Dataset: A simulated
dataset generated based on a random correlation matrix
49 Demonstration of ALADINS Simulation
50 RESULTS AND DISCUSSIONS
1. Discovery Component
2. Data Representation Component
3. Caching Component
 Analysis of Discovery Component
51

A series of experiments were conducted to analyze the impact of individual sub-

components on latency and hit ratio
Latency is the time difference when a discovery request is floated and the time
when a response is received
Hit Ratio is the ratio of discovery request locally satisfied to the total discovery
request issued by a node
Summary of the results are:

Next slides compare ALADIN with MoGATU

52 Latency Analysis of Discovery Component
53 Hit Ratio Analysis of Discovery Component
 54 RESULTS AND DISCUSSIONS
1. Discovery Component
2. Data Representation Component
3. Caching Component

54
 Growth of ontology size
during simulation is linear

55 Scalability of Data Representation Component

56 Latency Analysis of Data Representation Component
 57 RESULTS AND DISCUSSIONS
1. Discovery Component
2. Data Representation Component
3. Caching Component

57
58 Hit Ratio of Cache Replacement
59 Hit Ratio of Consistency Manager
 Research Contributions
60

We present a scalable, network layer, association rules based data

and information management framework for MANET
This research substantiates that:
the use of a multi-level schema gives rise to improvement in scalability
of data management operations
a network layer discovery approach leads to improvement in latency
and robustness of data management operations
the correlation patterns present in the request sessions can be
exploited in predictive fashion to ameliorate the discovery and caching
process
It is the first work(for MANET) exploring the use of :
data correlations
multi-level data representation
 Future Work
61

Improvement in discovery component

Proactive Component
Smart broadcasting based on available battery, network congestion
Reactive Component
Implementation for other reactive routing protocols like TORA, ABR, OLSR etc.
Association Rules Mining are inefficient in terms of memory consumption
New approximate algorithms can be devised
Caching Component
Based on proposed correlation technique, extensions of other caching
algorithms (MFU, Profile Driven caching)
Log database
Can it be used for transaction management, query optimization etc.?
Further work/evaluation of replication, query optimization and security
 References
62

Alfano, R. and A. Manzalini (2012). Method and System for Data Managment in Communication
Networks. U. Patents. US, Telcom Italia S.p.a., Milan (IT). US 8,165,130 B2.
Arias-Torres, D. and J. A. Garca-Macias (2004). Service Discovery in Mobile Ad-hoc Networks by
Extending the AODV Protocol. 2nd Mobile Computing Workshop(ENC' 04. 2004), Mexico.
Bharadvaj, H., A. Joshi and S. Auephanwiriyakyl (1998). An Active Transcoding Proxy to
SupportMobile Web Access. IEEE Symposium on Reliable Distributed Systems, Dana Point,
California.
Chand, N., R. C. Joshi and M. Misra (2007). "Cooperative caching in mobile ad hoc networks
based on data utility." Mobile Information Systems 3(1): 19-37.
Chen, H., F. Perich, T. Finin and A. Joshi (2004). SOUPA: Standard ontology for ubiquitous and
pervasive applications. The First Annual International Conference on Mobile and Ubiquitous
Systems: Networking and Services (MOBIQUITOUS 2004), Boston, Massachusetts, USA
Fatima, N. S. and P. S. A. Khader (2012). "Efficient Data Accessibility using TTL Based Caching
Strategy in Manet." European Journal of Scientific Research 69(1): 21-32.
Fontijn, W. and P. Boncz (2004). AmbientDB: P2P Data Management Middleware for Ambient
Intelligence. Pervasive Computing and Communications Workshops, 2004, Orlando, Florida,
USA.
 References (contd)
63

Gryazin, E. A. (2006). Service Discovery in Bluetooth. Helsinki, Finland, Group for Robotics and Virtual Reality.
Department of Computer Science. Helsinki University of Technology: 1-4.
Guttman, E., C. Perkins, J. Veizades and M. Day (1999). Service location protocol, version 2, RFC 2608, June
1999.
Han, J., J. Pei, Y. Yin and R. Mao (2004). "Mining Frequent Patterns without Candidate Generation: A Frequent-
Pattern Tree Approach." Data Mining and Knowledge Discovery 8: 53-87.
Hara, T. (2001). Effective replica allocation in ad hoc networks for improving data accessibility. Twentieth
Annual Joint Conference of the IEEE Computer and Communications Societies, Anchorage, Alaska, United
States.
Hara, T., N. Murakami and S. Nishio (2004). "Replica allocation for correlated data items in ad hoc sensor
networks." ACM Sigmod Record 33(1): 38-43.
Jayapal, C. and S. Vembu (2011). "Adaptive Service Discovery Protocol for Mobile Ad Hoc Networks." European
Journal of Scientific Research 49(1): 6-17.
Karlsson, J. (2012). Routing Security in Mobile Ad-hoc Networks. Informing Science and Information
Technology Education, Montreal, Canada.
Martin, L. and I. Demeure (2008). Using structured and segmented data for improving data sharing on
MANETs. 8th international conference on New technologies in distributed systems, Lyon, France.
Perich, F., S. Avancha, D. Chakraborty, A. Joshi and Y. Yesha (2002). Profile Driven Data Management for
Pervasive Environments. 13th International Conference on Database and Expert Systems Applications
(DEXA 2002), Aix-en-Provence, France.
 References (contd)
64

Perich, F., S. Avancha, A. Joshi, Y. Yesha and K. Joshi (2001). Query routing and processing in mobile ad-hoc
environments. Baltimore County, University of Maryland: 1-17.
Perich, F., A. Joshi and R. Chirkova (2006). Data Management for Mobile Ad-Hoc Networks. Enabling
Technologies for Wireless E-Business. 1: 132-176.
Sanderson, N. C., K. S. Skjelsvik, O. V. Drugan, M. Puar, V. Goebel, E. Munthe-Kaas and T. Plagemann (2007).
Developing Mobile Middleware-An Analysis of Rescue and Emergency Operations, Department of
Informatics, University of Oslo.
Shahram Ghandeharizadeh, A. H., Bhaskar Krishnamachari, Franois Bar, Todd Richmond (2006). Data
Management Techniques for Continuous Media In Ad-Hoc Networks of Wireless Devices. Collection of
Encyclopedia of Multimedia: 136-140.
Taneja, S. and A. Kush (2012). "Intrusion Detection Protocol for Adhoc Networks." Network and Complex
Systems 2(2): 22-29.
Xiong, W. A. and B. Tang (2011). "A Secure and Highly Efficient Key Management Scheme for MANET."
Advances on Information Sciences and Service Sciences 3(2): 12-22.
Yang, B. and A. R. Hurson (2007). "Semantic-Aware and QoS-Aware Image Caching in Ad Hoc Networks." IEEE
Transactions on Knowledge and Data Engineering 19(12): 1694-1707.
Zhong, T., B. Xu and O. Wolfson (2008). Disseminating real-time traffic information in vehicular ad-hoc
networks. Intelligent Vehicles Symposium, Eindhoven, Netherland.
Zink, M., K. Suh, Y. Gu and J. Kurose (2008). Watch global, cache local: YouTube network traffic at a campus
network-measurements and implications. 15th SPIE/ACM Multimedia Computing and Networking
(MMCN08), Santa Clara.