Top Banner
1 CHAPTER 1 INTRODUCTION 1.1 GENERAL In the last two decades, there has been a substantial improvement in the computing and network performance. The recent advances in technologies have enabled the availability of Personal Computers (PCs), workstations, Symmetric Multiple Processors (SMPs) and nomadic devices at a cheaper cost, making it possible to achieve high performance and high throughput computing. The low cost high performance computing systems motivated researchers to solve the resource and data intensive problems in a number of application domains. The availability of powerful computing resources and high speed networks allowed scientists to broaden their simulations and experiments to accommodate more parameters than ever before. The scientific applications often need a huge amount of computational and storage power. Hence, seamless access to such resources that are distributed geographically are needed. The power of computing and networking made it possible to share resources such as data from instruments, results of experiments with collaborators, and high performance computers around the globe almost instantaneously. Several attempts have been made to provide support for flexible and controlled sharing of various types of resources that are required to solve computationally intensive applications. For instance, the cluster computing paradigm fails to address the management
19

CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

Oct 18, 2020

Download

Documents

dariahiddleston
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

1

CHAPTER 1

INTRODUCTION

1.1 GENERAL

In the last two decades, there has been a substantial improvement in

the computing and network performance. The recent advances in technologies

have enabled the availability of Personal Computers (PCs), workstations,

Symmetric Multiple Processors (SMPs) and nomadic devices at a cheaper

cost, making it possible to achieve high performance and high throughput

computing. The low cost high performance computing systems motivated

researchers to solve the resource and data intensive problems in a number of

application domains. The availability of powerful computing resources and

high speed networks allowed scientists to broaden their simulations and

experiments to accommodate more parameters than ever before.

The scientific applications often need a huge amount of

computational and storage power. Hence, seamless access to such resources

that are distributed geographically are needed. The power of computing and

networking made it possible to share resources such as data from instruments,

results of experiments with collaborators, and high performance computers

around the globe almost instantaneously. Several attempts have been made to

provide support for flexible and controlled sharing of various types of

resources that are required to solve computationally intensive applications.

For instance, the cluster computing paradigm fails to address the management

Page 2: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

2

of heterogeneous resources, whereas distributed computing paradigm does not

support the management of resources belonging to diverse organizations.

To address this issue, an extended distributed computing

technology, termed as 'Grid', was coined by Ian Foster et al (2001 and 2002)

that support the aggregation of distributed computational resources that spans

beyond the organizational boundaries, and their coordinated use to meet the

requirements of advanced science and engineering. Grid can be distinguished

from conventional distributed computing by its focus on large scale resource

sharing, high performance, and computation / data intensive applications.

Grid supports researchers and scientists from diverse organizations to share

information, instruments, data, computation and storage resources

dynamically in a flexible and secure manner, thereby forming a ‘Virtual

Organization’(VO) to solve challenging applications.

To realize such an infrastructure, layered reference architecture

with a set of protocols was devised and is shown in the Figure 1.1. The

bottom layer is the resource (fabric) and top layer comprises of demanding e-

science applications.

Figure 1.1 Grid Protocol Architecture

Page 3: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

3

The middle three layers are also called as Middleware provides an

abstraction over the resources and relieves the user from the tenets of

management issues, like locating the suitable resource, job migration, security

and monitoring. Taking the analogy of Internet, which is a network of

networks, the grid is just an aggregation of clusters. The rationale of choosing

a cluster as the basic entity of the grid is that it can encapsulate a substantial

number of nodes and provide a single system image. Also the grid is

constructed by the contribution from various organizational resources. It is

easy to maintain the contributing organizations cluster than manage the

machines one by one.

The Resource and Connectivity layers contribute a set of services,

which performs the individual cluster management. The Collective layer on

top performs grid wide management, which is a super scheduler.

In addition to the middleware, to facilitate data centric applications

on the grid, domain specific functionalities can be complemented with the

middleware and classified as core services and higher level services as

depicted in Figure 1.2. The core services such as data discovery, job

submission and replication provide transparent access to the distributed data

and computation.

Figure 1.2 High Level Services for Data Centric Applications

Page 4: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

4

Higher level service includes resource brokering which selects

resources for a user based on his requirement and replica management service

which decides about the creation and distribution of multiple copies of the

same data item. These services can be constructed from the core middleware

services to complement the collective layer.

Grid is already being successfully used in many scientific

applications such as high-energy physics, bio-medicine, aerospace and earth

sciences and is continuing to evolve and expand. However, it is also the basis

and enabling technology for pervasive and utility computing due to its ability

of being open, highly heterogeneous and scalable. Yet, it has become essential

to support seamless computing and communication for mobile users, which

would offer flexibility and increased information availability.

At the same time, a modern computing technology arises with the

consumer electronic devices which is termed as Mobile Computing. Mobile

computing is an expansion of the traditional distributed computing, which

includes devices like Laptops with Wireless Local Area Networking (WLAN)

technologies, Mobile Phones and Personal Digital Assistants. This computing

model focuses on the requirement of providing access to information,

communication and services everywhere, anytime.

The rapid technology advancements made mobile devices more

powerful (in terms of computing) and ubiquitous (in terms of connectivity).

The evolution of 3G and 4G wireless technologies enhances the spectral

efficiency, to support high data rates and mobile broad band everywhere.

Nowadays, the wireless devices such as laptops are enriched with a

substantial amount of computing power equal to their static counterparts and

the ability to change the locations with the same IP address is also achieved.

These advances made the wireless devices to offer high performance

capabilities in a flexible manner seamlessly.

Page 5: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

5

Figure 1.3 depicts the common mobile application architecture

proposed by Microsoft patterns and Practices team (2009) to enable the

applications on mobile computing environment.

Figure 1.3 Mobile Application Architecture

In this architecture, the mobile devices are connected with the base stations using wireless communication medium and the connectivity extended with the mobile support infrastructure. The two main components of the architecture are

1. Application support layer

2. Infrastructure support layer.

This architecture enables the realization of data centric applications

on the mobile domain by having a separate layer called data layer. The data

specific provisions like replica management, replica selection and data access

utilities can be integrated with this data layer. Rapid advances in the

Page 6: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

6

increasing processing power of wireless devices, along with the wireless

technologies such as Wi-Fi, Wi -Max has made them increasingly capable of

participating in grid networks.

The combinations of these two computing models have the

potential to realize significant developments in the adoption of high

performance grid through mobile devices. As a result of several case studies

from various e-Science projects as discussed by Hey T and Trefethen A E

(2002), the mobile device needs the grid for computation and integration,

while the grid needs the mobile devices to interface with the physical world.

With the incorporation of third generation mobile networks and high

bandwidth wireless technologies, grid computing has moved from the

traditional parallel and distributed model to the mobility based model (Sanjay

Ahuja and Jack Myers 2006).

This paradigm shift has given rise to a new mechanism called

Mobile Grid (M-Grid). The Mobile Grid enables both the fixed and mobile

users to have access to both the fixed and mobile grid resources transparently

using the underlying technologies. The integration of the fixed and mobile

devices into the grid environment is presented in the Figure 1.4.

Figure 1.4 Mobile Grid Computing Model

Interface to both fixed and

mobile grid resources

Page 7: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

7

In the above mobile grid model, the two possible roles of mobile

devices are,

Mobile device as a Grid Interface

o Mobile devices transfer their tasks to the grid rather than

performing it themselves and monitor them being

executed on the grid. In this role, mobile devices act as

Grid Service consumers as shown in the Figure 1.5.

Figure 1.5 Mobile Devices as Grid Interface

Mobile device as a compute node in grid infrastructure

o This integration makes mobile devices aggregated into

the grid environment as resource providers for effective

computational distribution. In this, by forming Mobile

Dynamic Virtual Organization (MDVO) (Martin

Waldburger and Burkhard Stiller 2005) the wireless

devices participate in the grid as resources. The Figure

1.6 shows the mobile devices as resource provider in

grid.

Page 8: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

8

Figure 1.6 Mobile Devices as Grid Resources

While integrating mobile devices with the grid environment which

has a dynamic nature, the conventional grid middleware needs extension to

support heterogeneous device characteristics, disconnected operations,

resource optimization in resource constrained environment and energy

optimization. Also, the main objective of realizing mobile grid is running data

centric applications. In such applications, large amount of data will be stored,

accessed and processed. To enhance these operations, an effective high level

data management service like data replication can be utilized.

1.2 DATA REPLICATION IN GRID DOMAIN

Data replication is a well known strategy to improve the

performance and reliability of any distributed computing platforms. When the

users of a system are distributed over a resource sharing network such as grid,

keeping data at a single location can affect data access in three ways

Latency: The data access time varies with the distance and link

bandwidth of the user from the data storage, and it is subject to network

problems related to the network environment.

Grid Node

Page 9: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

9

Availability: Having a single data storage site is risky for critical

applications. When storage is temporarily unavailable (for faults or

maintenance reasons), or the storage site is not reachable due to network

problems, users do not have access to data.

Congestion: The single data storage site must sustain a potentially

high number of user’s requests. The hardware used for data storage can be

very expensive or fail to satisfy the user’s requests.

For these reasons, to facilitate the data centric applications on the

grid domain, data replication services seek to enhance the network traffic by

copying heavily accessed data to appropriate locations and managing them.

The replication mechanism attempts to determine when, where and what data

are to be replicated across the grid nodes that may be fixed or mobile. The

Figure 1.7 illustrates a replication environment.

Figure 1.7 A Simple Replication Scenario

In the Figure 1.7, Site 1, Site 2, …, Site n are distributed site

locations and connected through a middleware infrastructure. Data stored in a

file, File X, is stored in site 2and is replicated at all other sites. The benefits

Page 10: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

10

due to replication of files can be accessed at a cheaper cost i.e. closer to the

user and the file can be still accessed even if 3 out of the 4 sites are down.

Data replication increases the availability and reliability of the data. It also

reduces data access delay at the cost of data storage.

However, in a replicated environment different challenges must be addressed, and the following are some of the key points,

Data Consistency

o Maintaining uniformity among the distributed data

Maintenance Overhead

o Managing more number of copies makes it difficult to administrate in terms of storage and update propagation.

Lower Write performance

o Applications require more updates in a replicated environment; multiple copies may have to be updated.

Due to the above mentioned challenges, it is necessary to manage

the resources in a resource restricted (intermittent connectivity, battery power)

mobile grid environment in an optimized manner and the service provisions

should be reliable.

1.3 ADOPTION OF MOBILE GRID

The adoption of this new computing model research is already

accomplished in the areas of disaster handling, e-health and crisis

management. Some of the currently existing mobile grid projects are given

below,

AKOGRIMO - Developed mobile collaborative business grids for

e-health, e-learning and disaster management. (www.akogrimo.org)

Page 11: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

11

WiGiT - A middle ware extension layer called Edgeware to

facilitate collaboration among grid, cloud and mobile networks.

(www.wigit.ischool.syr.edu)

Garuda Grid - Provides Garuda Access Portal 2.1 with mobile

services, supporting access and monitor the grid resources through mobile

phones. (http://portal.garudaindia.in/gap2/gap2)

The motivation and contributions of the present research, to extend

the middleware capabilities to a mobile grid and enhance data centric

applications on it are discussed in the next section.

1.4 MOTIVATION AND CONTRIBUTIONS

The integration of the grid and mobile computing models provide a

remarkable development ie, the adoption of Wireless Cooperative Clusters

(WCC) into the high performance grid. Grid can include these clusters as task

execution environment by forming MDVO.

The core requirements to realize such an environment can be

grouped into two levels of abstractions as mentioned below,

Infrastructure Management

o Managing intermittent connectivity

o Autonomic configuration

o Energy optimization

o Resource optimization

Data Management

o Availability ( Replication)

o Reliability ( Fault tolerant)

Page 12: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

12

o Reliable task achievement

o Maintainability

1.4.1 Infrastructure Management

In a mobile environment, the available resources comprise of static

fixed nodes and dynamic mobile nodes. To accommodate the dynamic nature

of the mobile grid environment such as link quality and mobility of nodes, the

conventional middleware components does not seem to be flexible and

scalable. The formation of a mobile grid requires additional functionalities to

be added at the infrastructure level in a flexible and open way which are listed

below,

Resource management and optimization

Mobility management

Asymmetry connectivity

Changing loads on the participating node

Energy optimization and

Formation of MDVO

The volatile and dynamic mobile grid environment requires the use

of sophisticated mechanisms for resource discovery and selection. The

selection of resources that meet the time and cost constraints should be

imposed by the adapting mechanism. Additional match making parameters

like resource accessibility, system workload and network performance have to

be considered by the approach.

Constraints that would complicate the mobile grid job scheduling

are the changing devices connection status and their physical locations.

Hence, an optimization criterion for job scheduling mechanism should

Page 13: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

13

consider not only the cost and performance of each resource in terms of task

execution, but also the current availability of the reliable resource in a

MDVO. One such network characteristic to achieve reliability in the mobile

environment is the Signal – to – Noise Ratio (SNR). SNR directly impacts the

performance of a wireless LAN connection. A higher SNR value (in dB)

means that the signal strength is stronger in relation to the noise levels, which

allows higher data rates and fewer re-transmissions.

The bandwidth, storage, computational and energy resources spent

for task execution, is wasted if the assigned task is aborted upon

disconnection. It is stressed that the approaches used for reliable task

execution, should address the problem of intermittent connectivity. Most of

the approaches in this direction are not mutually exclusive. Building

connectivity profiles for each node may assist with an intelligent scheduler

and such intelligence can be achieved using agent based technology.

To make grid functionalities feasible among the group of mobile

devices, energy efficiency is one of the major concern for their

implementation on each node. To find a technique to reduce the power

consumption for communication and storage is an issue. It is mandatory to

provide a mechanism, to periodically exchange the power level of nodes, so

that the node with the highest power can be considered to hold the services, to

improve the availability.

1.4.2 Data Management

The idea of integrating mobile devices with the back end parallel

computing cluster will be essential for different classes of parallel

applications such as image rendering on battlefield, surveillance,

reconnaissance and other classes of military applications. In these scenarios,

attack task forces can form a mobile grid network to accomplish the mission

Page 14: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

14

(also called mobile tactical networks Yang Zhang et al (2011)). As the

mission progresses, a large amount of information must be shared and

processed among the soldiers and squads by their wireless devices. In such

cases the data replication services are a vital requirement to facilitate the

execution of data centric applications on the mobile grid.

The key challenge in mobile grid with respect to data service is the

effective replication of data over mobile nodes. To improve and maintain the

overall throughput of grid jobs that access files, the environment has to

provide an optimal distribution and replication of data files. With a number

of replicas available, a client site can access the file with minimal distance.

However, maintaining multiple copies in a resource constrained mobile grid

system is expensive. Therefore, the number of replica should be bounded to

the optimum number of copies.

Most of the replica management procedure supports only

immutable file operations in the conventional grid implementation. It is

mandatory to assist mutable file transactions to realize the grid in the

commercial domain. The new computing model like Mobile Grid should

consider both the read and update transactions to effectively manage the

replication in this domain.

Due to the dynamic characteristics of mobile grid such as mobility

of the nodes, the node which holds replica currently may not be the best site

to facilitate the access request. Therefore, relocation is to be considered if the

performance is to be maintained. While relocating a replica, user access

patterns may be considered to minimize the access cost. Based on the

accumulated Read / Write statistics of a node, it can be selected to hold the

replica.

Page 15: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

15

In order to handle the intermittent connectivity among the mobile

devices, it is necessary to consider the signal quality in terms of SNR. During

the selection of the nodes to hold the replica, the node which has the higher

SNR can be considered. Hence, one can prevent data replication at nodes

which has the tendency towards connection failure. Hence, the reliability and

stability of the network can be maximized.

Apart from considering mechanisms to find the location of replica,

relocation of replica, consistency maintenance and infrastructure for mobile

grid, use of good replica placement strategies is necessary to optimize the

number of replicas and to improve the performance of replica aware services.

Additional challenges to be considered with regard to resource

management are optimized resource utilization, energy optimization,

autonomic configuration and mobility management. It requires a well defined

fault tolerant provision to improve the performance in terms of scalability and

reliability.

In conclusion the management of data replication has to be

achieved in an effective manner in the mobile grid environment requiring

additional efforts. It has to cover all the aspects of wired and wireless grid

characteristics. Considering only the static characteristics will not be a

solution for replica management. In the mobile grid a lot of dynamic

characteristics such as topology management, transmission capability,

handling of flash request with minimum resources and signal strength (SNR)

of the environment must be considered. In addition, agent based intelligence

can be used to support the replication process.

The problems considered in the proposed research are depicted in

the Figure 1.8.

Page 16: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

16

Replication Support

Infrastructure Support

Figure 1.8 Architecture of the Proposed Work

The present research work presented in this thesis has addressed a

multi agent based framework to enhance the middleware to support mobility

and a replica management mechanism to support data services on the mobile

grid. The proposed approach is enhanced with consistency management

among the replications and fault tolerant provisions to manage the failure.

The work considers the dynamic characteristics such as Signal quality,

Storage and disconnection.

In brief, the contributions of the research are summarized as

follows,

User Applications

Grid Middleware

Mobility support agent based framework

High level Data Services Optimized Replication Manager

Grid Portal

Consistency Maintenance Support

Fault Tolerant Support

Static Resources Dynamic Resources

Page 17: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

17

1. Proposed an optimized multi-site replica placement mechanism based on Signal-to-Noise ratio, which is intended to find the appropriate and reliable candidate nodes to place the replica. The proposed algorithm finds, deletes, and relocates the candidate node, based on the read and update frequency, and communication cost along with the threshold Signal –to– Noise ratio. Also, this mechanism will select only minimum number of locations to place the replica which can satisfy large spike of unpredictable requests. Hence, maintenance becomes easy with good availability.

2. Existing middleware mechanisms supports the integration of mobile devices as resource consumers. However, it is necessary to build a mobility aware framework to allow the wireless devices to participate in the grid as resource provider by pooling the resources. To realize this, a multi-agent based replication framework is proposed which provides a comprehensive infrastructure for improving data availability with a small number of replicas. The functional components that are required to support mobility such as system state monitor, context analyzer, strategy manager, localization manager, replica manager and consistency manager are proposed. The architecture is enhanced with multi agents, to intelligently manage the replica and ensure consistency in the resource constrained environment.

3. In a replicated environment, it is mandatory to maintain the consistency among the multiple copies of the data distributed in multiple locations. In order to achieve that, a consistency management approach which uses the top down update propagation method is proposed. In this context, the

Page 18: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

18

base station manages the update conflict by maintaining the mutual exclusion to access the data resources.

4. The ubiquitous distributed computing environment should maintain the stability of the environment and it should be continuous to operate despite the failures. A fault tolerance algorithm, which exhibits a well defined failure behavior, is proposed. This approach triggers in case of node failure in the environment.

5. The simulation environment is adopted to evaluate the system performance. Different scenarios have been designed and analyzed based on the signal-to-noise ratio, communication cost, data access cost, mobility information, load information and battery information.

1.5 ORGANIZATION OF THE THESIS

The chapters are organized as follows. The Chapter 2 presents the

existing research contributions, in terms of mobile grid middleware support

for various applications. Further research issues towards data replication with

mobility support and adaptive data management on mobile grid environment

are studied. The different middleware are compared for their service support

such as resource allocation support, dynamic environment support and mobile

device execution support. In addition various replication mechanisms for

mobile grid environment are analyzed with their advantages and limitations to

understand the strategies.

In Chapter 3, the extension at higher level data service layer is

explained. An optimized multi site replica placement algorithm is elaborated.

The enhanced method works with three tests namely, expansion test,

contraction test and switch test. The proposed placement mechanism is based

Page 19: CHAPTER 1 INTRODUCTION - Shodhgangashodhganga.inflibnet.ac.in/bitstream/10603/24255/6/06_chapter1.pdf · have enabled the availability of Personal Computers (PCs), workstations, Symmetric

19

on the Signal – to – Noise ratio for mobile grids, which eliminates some of the

drawbacks of the existing systems. Multiple nodes for replica placement are

selected in a single iteration, catering the needs of sudden burst of requests.

The internal components for infrastructure level extension are

elaborated in Chapter 4. The implementation of the proposed replication

framework using agent technology to support the mobile grid is presented.

Two logical components are provided at the base station level and the mobile

host level. Design considerations of the proposed functional components that

are required to support mobility are discussed. To support replica

management effectively with this framework, three types of agents namely,

Base Station agent (BS agent), Node agent (N agent) and Update agent

(U agent) are realized. The experimental results demonstrate the benefits of

the usage of agents in the proposed framework.

In a dynamic replicated environment, maintaining consistency and

failure handling are important aspects. The Chapter 5 discusses the update

propagation mechanism to maintain consistency among the replicas. Different

agent states have been discussed and the interactions between these states are

elaborated. The proposed approach uses ROWA (Read One Write All) model.

The deferred update procedure is used to ensure replica consistency after each

write operation and resolving update conflicts. Along with the consistency

maintenance a fault tolerant algorithm is proposed and explained in Chapter 6.

To maintain the stability of the system in case of node failures, the algorithm

considers the mobility, battery and load information for selecting the node to

place the replica. Chapter 7 concludes by reviewing the present research and

discussing the future directions of the mobile grid computing.