Fault-tolerant Mobile Agent-based Monitoring Mechanism for Highly Dynamic Distributed Networks

grapedraughtSoftware and s/w Development

Dec 2, 2013 (3 years and 8 months ago)

79 views


Fault
-
tolerant Mobile Agent
-
based Monitoring
Mechanism for Highly Dynamic Distributed
Networks


IJCSI International Journal of Computer Science Issues, Vol. 7,
Issue 3, No 3, May 2010


1. Abstract
:


A

certain number of mobile agent
-
based monitoring mechan
isms

have actively been developed to monitor large
-
scale and

dynamic
distributed networked systems adaptively and

efficiently. Among
them, some mechanisms attempt to adapt to

dynamic changes in
various aspects such as network traffic

patterns, resource add
ition
and deletion, network topology and so

on. However, failures of
some domain managers are very critical

to providing correct, real
-
time and efficient monitoring

functionality in a large
-
scale mobile
agent
-
based distributed

monitoring system. In this pa
per, we
present a novel
fault tolerance

mechanism to have the following
advantageous

features appropriate for large
-
scale and dynamic
hierarchical

mobile agent
-
based monitoring organizations. It
supports fast

failure detection functionality with low failur
e
-
free
overhead by

each domain manager transmitting heart
-
beat
messages to its

immediate higher
-
level manager. Also, it
minimizes the number

of non
-
faulty monitoring managers affected
by failures of

domain managers. Moreover, it allows consistent
failure

d
etection actions to be performed continuously in case of
agent

creation, migration and termination, and is able to execute

consistent takeover actions even in concurrent failures of domain

managers.




In order to determine whether exis
ting monitoring systems can
satisfy the requirements imposed by future “networked” systems
we need to identify first these requirements. An important
difference between present day and future networking is that in the
former case topology is considered rel
atively static while this
became an unsafe assumption for future networks. New models of
dynamic networks are constraining with the situation in which the
networks topology was mainly modified as a result of careful
planning. Examples in which logical topo
logy layering can be
dynamically constructed in real
-
time are the following:



Dynamically re
-
configurable networks



Active networks



Dynamic Virtual Private Networks



Mobile and survivable networks



Re
-
configurable cellular networks



2. EXISTING

SYSTEM


The cu
rrently available system can be categorized as

o

Static centralized monitoring

o

Static decentralized monitoring

Hereby the description and the functions of the above systems
would be described such that it would explain the need why we
have suggested and imp
lemented our system.


2.1 STATIC CENTRALISED MONITORING



In this case there is a single monitoring station with which all
the monitored systems communicate directly. The monitoring
station is in charge of collecting, aggregating and processing raw
network

data.



This model is widely used to manage relatively static, small
networks using (simple network management protocol) SNMP.
The model has been criticized for its limited responsiveness,
accuracy and lack of scalability. The concentration of management

intelligence in a single point results in processing and
communication bottlenecks, limiting the number of elements that
can be monitored and the rate at which information can be
gathered. Furthermore SNMP favors a polling approach which
limits the abilit
y to track problems in a timely manner while
requiring management traffic even if no significant change has
occurred.


To overcome the shortcomings of polling, the alternate
technique of event reporting may be used. With event reporting,
the monit
ored systems take he initiative to inform the manager
according to pre
-
determined rules set by the manager. Event
reports are generated within the monitored systems either
periodically or as and when an critical event occurs.


Periodic reporting pr
ovides the manager with status
information in a summarized manner and is more efficient than
requesting the same information via polling. On the other hand
alarm reporting is used for detecting problems as soon as they
occur. The problem with alarm reporti
ng is that the types of alarms
need to be thought
-
out in advance, standardized and supported by
vendors. Event reporting requires an increased level of intelligence
in the monitored systems.



Typical systems employ both polling and event reporting
althoug
h in practice the telecommunications systems rely more on
event reporting and SNMP
-
based management systems



2.2 STATIC DECENTRALIZED MONITORING



One way to increase performance and scalability is to
adopt a hierarchical management architectu
re which uses multiple
systems with one system acting as a main monitoring station and
the others working as area monitors. Hierarchical monitoring is
used in Telecommunications Management Network (TMN). In
context of SNMP, simple monitoring and statistica
l probes can be
introduced using RMON, which is equivalent to an area monitor
that collects monitoring information about a number of elements
within a sub
-
network. More recently, other forms of
decentralization based on distributed object technologies such

as
CORBA and JAVA RMI have become popular in management. An
extensive review of management paradigms and technologies can
be found.


The common denominator of the above approaches in the
adoption of simple, pre
-
defined functionality that can actually be
d
ecentralized is restrained to operations such as low
-
level filtering
of monitoring data, generation of alarms on the basis of simple
conditions, and collection of rudimentary statistical information. In
addition, these decentralized area monitors operate i
n pre
-
defined
network locations, which mean that they cannot easily adapt to
network changes. Therefore, conventional static decentralized
schemes, despite coping with the scalability problem to a certain
extent, inherit the other problems of centralized m
anagement and
cannot easily cope with frequently changing, dynamic
environments.

So far we have been discussing the operations and problems
concerned with the static systems. Hence to overcome the cons of
static systems, we are looking into the dynamic as
pect of
methodologies of which the programmable decentralized
monitoring poses some shortcomings not satisfying the
requirements.


Monitoring
Station

Internet

STATIC CENTRALISED MONITORING

Figure 1.1

STATIC DECENRALISED MONITORING

Figure 1.2

Monitoring
Station

Internet






Monitoring
Station

Internet

Dynamic Decentr
alized Monitoring

Figure 1.3

MOBILE AGENT

HOST

AREA MONITOR

2.3 PROGRAMMABLE DECENTRALISED MONITORING



When we talk about d
ynamic or programmable, it deploys
new management logic ‘when’ and ‘where’ is needed without
having to predefine the logic. With distributed object technologies,
the management logic can only be modified through software re
-
installation.



The first propos
al to support remote programmability was
introduced with the use of mobile code in network management,
signaling a paradigm shift from static to dynamic management.
The basic underlying principle is that new management functions
can be dynamically introduc
ed to a managed node as required. The
manager uses the protocol to ‘push’ new code down to a managed
node; management routines are executed locally rather than
centrally at management station. Therefore, a mechanism to
decentralize management processing an
d to re
-
program managed
node capability.



Managed nodes were relatively simple in terms of processing
power and there was no uniformity in processing environments. It
is the increase in processing power and with the advent of Java that
the paradigm has be
come a viable solution. Java’s object
serialization makes it easy to migrate code whilst Java
-
RMI
provides for simple communication between distributed objects.


The single
-
hop mobility mechanism, despite being extremely
useful as a mechanism for flexible

and dynamical remote
programmability, is still a relatively static mechanism since it is
only used to deploy management logic at start up time. The
decision of ‘when’ and ‘where’ to deploy management logic is still
taken by a centralized management statio
n based on a static
network view. Because a MA is conceived as a dynamically
deployable piece of code rather than being free to roam the
network, full code mobility is not exploited to provide run
-
time
adaptation. Therefore, the single
-
hop mobility mechani
sm does not
fully satisfy the requirement of large
-
scale, highly dynamic
networked systems



3. PROPOSED

SYSTEM


The system that we have proposed is based on the dynamic or
programmable methodology. It satisfies the requirements to
monitor the performance
for a large scale network efficiently and
effectively. It is known as the active distributed monitoring.


3.1 ACTIVE DISTRIBUTED MONITORING



The possible advantages of using agent mobility for network
management. Some of the pros are reduction of network
traffic,
increased responsiveness and robustness.


The problem addressed here describes how to exploit agent
multiple
-
hop mobility to build a distributed monitoring system
which reconfigures itself as the status of the monitored system
changes. Reconfigur
ability is an essential requirement if the status
of the monitored system is dynamic and transient. We have seen
that with distributed objects and single hop mobility we can only
realize a relatively static monitoring system that may or may not be
optimize
d on the basics of the initial status of the monitored
system. As the latter evolves, the distributed monitoring logic may
have to be relocated in order to maintain optimality
-

i.e. When
MAs are used as adaptive area monitors their optimal locations
depend

on the status of the network which may vary considerably
in highly dynamic environments.



The system is decentralized because the monitored system is
partitioned and separate agents are dynamically assigned to disjoint
partitions. Network partitioning is

computed in a distributed
fashion by the agent system. Finally, because agents are capable of
sensing the network status and migrate at run
-
time to maintain
location optimality, the system is “active” or adaptive. Such a
system exploits not only multiple
-
hop mobility but also agent
autonomy (each agent contains the logic to independently decide
when and where to migrate) and agent cloning i.e. the ability of an
agent to create and dispatch.





4. SYSTEM

REQUIREMENT

4.1 HARDWARE SPECIFICATION

Processor Ty
pe:
Pentium
-
IV

Speed : 1.2 GHZ

Ram : 512MB RAM


Hard disk : 40GB HD




4.2 SOFTWARE SPECIFICATION

Operating System

: Win2000/xp


Language


:JAVA, SWING, RMI

Protocol



:TCP/IP

Tool Used



:
-

Eclipse.