Application of Bayesian Networks for Autonomic NGN Management

placecornersdeceitΤεχνίτη Νοημοσύνη και Ρομποτική

7 Νοε 2013 (πριν από 3 χρόνια και 5 μήνες)

53 εμφανίσεις

Application of Bayesian Networks for
Autonomic NGN Management
Presented by
Abul Bashar
Outline
￿￿
Introduction
￿
Research Challenges
￿
Proposed Approach
￿
Implementation Details
Introduction
W
I
M
A
X
D
S
L
G
S
M
W
L
A
N
P
S
T
N
NGN
￿
Communication networks
Earlier: Stand-alone networks
Present: Converged network, NGN
￿
Network Management Systems
Ensures smooth running of network/services
Controls operational cost
Increases availability and QoS
NGN Architecture
Challenges and Solutions
What is a Bayesian Network?
Proposed Framework
o
nfig Data
Mgmt.
Data
Intelligent
Control Decisions
Network Management
System
SLA
Monitor
MIB
Monitor
Decision Support
System
BN
Policy
Engine
M
gmt.Data
Feature
Selection
Discretise
Variables
Structural
Learning
Parameter
Learning
Increment
Training Size
Implementation Details
Theoretical Framework
￿
To consider the desired objectives of the NGN (e.g. QoS metrics)

Delay

Throughput

Packet loss

Jitter
￿
To apply the following algorithms to realise the BN-based solution

Structural learning: PC and NPC algorithms

Parameter learning: EM algorithm

Sequential learning: Adaptation algorithm
Case Study (Admission Control )
Network Topology
BN Model
1
Case Study (Energy Aware NM )

Comparative Study (BN v/s NN )
Framework for a single router
BN Model
Distributed BN scenario
Multiple router scenario
Distributed BN Model (MEBN)
Internship at British Telecom
￿
Motivation

Successful implementation of BN in simulated scenarios

Need for validation of the proposed solution

Reply to the criticism of lack of real data and solution practicality

Availability of real data from BT’s 21CN monitoring systems

Opportunity to work with a reputed telecom market leader
￿
Research problem
Future work : Meshed Network
Future work : Mathematical Model
We define an entity to be our BN model for a single edge router pair (namely
xx
ERIR

, where
),...2,1Nx
=
.
For generality we assume the number of entities to be
N
, named
1
Y
to
n
Y
. Then the joint probability of the
overall domain (which is composed of multiple entities) will be:


),....,|(),....,|()(),...,,(
12112121−
×
×
=
NNN
yyyyPyyPyPyyyP




=
j
jj
yyyP),...,|(
11
, where
Nj,...,2,1
=
(1)

where from previous Eq.,


=
i
ijijj
XParentsxPyP))(|()(

where
ij
X
is the
th
i
node of BN corresponding to the
th
j
entity in the MEBN.
Finally applying the reasoning that, not every entity is dependent on another, we can define the joint probability as:



Conclusion
￿
BN have been shown to effectively model network
scenarios and achieve desired management goals
￿
Case studies : CAC, Energy-aware traffic
engineering, comparison with NN, distributed BN
Publications
[1]
A. Bashar, G.P. Parr, S.I. McClean, B.W. Scotney, D. Nauck, "Knowledge Discovery using Bayesian Network Framework for
Intelligent Telecommunication Network Management," in Proc. of Springer LNCS LNAI series,
4th International Conference on
Knowledge Science, Engineering &Management (KSEM 2010),
Belfast, UK, pp. 518-529, 1-3 Sep, 2010.
[2]
A. Bashar, G.P. Parr, S.I. McClean, B.W. Scotney, M. Subramanian, S.K Chaudhari, T.A. Gonsalves, "Employing Bayesian Belief
Networks for energy efficient Network Management," in Proc. of
16
th
IEEE National Conference onCommunications (NCC
2010)
,
IIT Madras, pp.1-5, 29-31 Jan. 2010.
[3]
A. Bashar, G.P. Parr, S.I. McClean, B.W. Scotney, D. Nauck, "Machine Learning based Call Admission Control Approaches: A
Comparative Study," in Proc. of
6
th
IEEE/IFIP International Conference on Network and Service Management (CNSM 2010)
,
Niagara Falls, Canada, pp. 431-434, 25-29 Oct, 2010.
[4]
A. Bashar, G.P. Parr, S.I. McClean, B.W. Scotney, D. Nauck, "A Novel Distributed Call Admission Control Solution based on
Machine Learning Approach," in Proc. of
12th IEEE/IFIP International Symposium on Integrated Network Management(IM
2011),
Dublin, 23-27 May, 2011.
(To appear)
[5]
A. Bashar, G.P. Parr, S.I. McClean, B.W. Scotney, D. Nauck, "Learning
-
based Call Admission Control Framework for QoS
Conference papers
THANK YOU