Semi-Stateful QoS Models in
The existing literature on
Quality-of-Service (QoS) in ad hoc networks can be broadly
categorized into stateful QoS
and stateless QoS architectures. Providing stateful QoS in MANETs is not significant in the presence of low traffic
and high mobility and it is too costly due to scarce resource in the network.
Stateless QoS, on the other hand, faces the problem
of stale QoS bandwidth or delay information since
there is no mechanism to keep the information of the flows status. However,
there has been a little work investigating the semi-stateful
QoS model. Therefore, this study aims to further
investigate the semi-stateful QoS
model by deploying a new resource reservation mechanism to the stateless QoS in order to provide QoS
guarantees in MANETs. Furthermore, this research also
investigates how adaptive QoS models can further be
developed in order to achieve QoS guarantees in
Broadcast
Algorithms in Ad Hoc
An Ad-hoc network is a collection of mobile nodes forming a temporary
network without any centralized administration, so nodes must cooperate to route
a packet. Broadcast service is a fundamental operation in ad hoc networks for
different application. It is often necessary in MANET routing protocols to
establish route to particular host, paging a particular host, and sending an
alarm signal. However, broadcasting by native flooding is usually very costly
and will result in serious redundancy, contention, and collision. The goal of
this research is the development and analysis of more efficient broadcast
methods that can minimize the number of retransmissions while ensuring high
delivery of broadcast packets.
Performance Analysis of Routing Algorithms for MANETs
Mobile ad
hoc networks (MANETs) have become the focus of
intense research in recent times due to the advent of affordable mobile devices
and the wide applicability of such networks to military and civil purposes.
Nodes participating in such a network can be highly mobile and so exhibit
unexpected connectivity changes that may radically change the topology of the
network. This sort of environment has prompted the development of new routing
algorithms that will address the challenges of optimal route discovery, goodput increase and effective scalability. Several ad hoc
routing algorithms have been proposed to take up those challenges. These can be
classified as proactive or reactive. Proactive routing algorithms discover
routes between source and destinations whether there is use for those routes or
not. Alternatively, reactive or "on demand" routing algorithms only
try to discover a route once it is needed. In this research, we conduct
an extensive performance comparison of some well-known routing algorithms with
the aim of identifying of identifying any performance deficiencies and propose
practical methods to overcome them.
A Bandwidth Efficient Routing Protocol for MANETs
MANETs have dynamic topologies by nature,
and have some limitations such as limited bandwidth and limited power. A
routing protocol that is able to deal efficiently with its dynamic nature and
its natural limitations is crucial to achieving good performance levels in MANETs. The goal of this research project is to develop a
bandwidth efficient routing protocol that can exploit with the dynamic nature
of MANETs to establish and maintain correct routes
between any source and destination, while reducing the use of limited resources
such as bandwidth and battery life.
Performance Evaluation
of Wormhole Networks with Finite Buffers and Multiple Virtual Channels
Analytical models for wormhole routed interconnection networks have been
widely reported in the literature. However, almost, all of these models have
assumed a one flit buffer in each networks node. Nevertheless, routers in
practical multicomputers are often equipped with
large (and finite) buffer size to hold transiting messages in order to smooth
out congestion. This research aims to develop analytical models for
wormhole-routed k-ary n-cube networks with finite
size buffers. Currant interconnection
networks exploit the concept of virtual channels not only to avoid deadlock
situations but also to better utilize the physical bandwidth. Almost all models
that have considered virtual channels have used the method proposed by Dally to
capture the effects of virtual channels multiplexing. However, it has been
revealed that Dally’s method, is accurate only
under low traffic. This is due to the statistical similarities of different
queuing systems under low traffic rates. The accuracy of the method degrades
under moderate and high traffic conditions. This study aims to further investigate
the effects of virtual channel multiplexing on network performance using the
analytical approach.
Past Research
Scalable
Algorithms for Broadcast and Multicast in Mesh Networks
The mesh has been recognized as one of the most suitable networks for
large-scale multicomputers due to its ease of
implementation, high scalability, and ability to exploit communication
locality. In this work, a new routing approach, the Coded Path Routing (CPR),
will be used as a new framework for designing efficient collective algorithms,
e.g. broadcast and multicast, for the mesh. In
the first stage of this research, new broadcast
algorithms for the 2 and 3-D mesh are proposed. These algorithms can
exploit advanced system features, including the multiport model, routing adaptivity,
and virtual channels. An extensive comparative study will be performed using
analytical and simulation models against existing algorithms that use the
common unicast and multidestination-based
approaches. In the second stage, new algorithms
for multicast in the 2D and 3D mesh will be proposed and compared to the
previous well-known algorithm.
Fault-Tolerant Replication Algorithms in Distributed Systems
Replication is a key to providing good performance, high availability,
and fault tolerance. In our research we study fault tolerant dynamic data
replication in distributed systems. We aim at designing data replication
algorithms, developing fault tolerant techniques capable of providing high
availability in presence of system failures, and analyzing the performance of
these algorithms and techniques. We also study the application of such
algorithms in areas like the Internet, Multimedia servers, and mobile
computers.
Performance Modelling and Analysis of Routing Algorithms with Deadlock
Recovery Strategies
Deadlock recovery as a viable alternative to deadlock avoidance has
recently gained consideration in the scientific community. It has been shown
that deadlocks are quite rare except when the network is close to saturation.
Thus the hardware dedicated for deadlock avoidance is not necessary most of the
time. This consideration has motivated several researchers to propose different
routing algorithms based on regressive or progressive deadlock recovery
methods. Except studies that have resorted to simulation to evaluate the
performance benefits of the deadlock recovery routing algorithms, there has
been hardly any work that describes analytical models for fully adaptive
routing algorithms with deadlock recovery. This research is intended to
contribute towards filling this gap by presenting new analytical models for
fully adaptive routing based on deadlock recovery.
Parallel
Algorithms in Hierarchical and Optical Networks
In this research, we propose a new hierarchical interconnection network,
referred to as the arrangement-star, deriving many of its topological and
performance properties from its constitutes the
well-known star and arrangement graphs. We construct a framework for the
development of efficient algorithms on this new network. We extend the
algorithmic framework to another recently-proposed
hierarchical topologies including the optical transpose system networks (or
OTIS networks). Our research aims to show that our algorithmic framework
enables hierarchical networks to outperform their flat counterparts, including
mesh, k-ary n-cubes, and hypercubes.
Performance Modelling/Evaluation of Wormhole-Routed
Networks under Non-Uniform Traffic
All the performance models proposed for direct wormhole-routed networks
have used uniform message traffic and most of them assume deterministic
routing. We are studying the impact of the non-uniformities in message traffic
on the performance of wormhole-routed k-ary n-cubes
using analytical models. Our focus is adaptive routing,
as such routing algorithms are getting popular in practice. We have been
studying three well-known non-uniform traffic patterns that have already been
examined using simulation. They include traffic generated by hot-spots,
matrix-transpose and digit/bit reversal.
Probability-Based Fault-Tolerant Routing Algorithms
A routing
algorithm specifies how a message selects a path to cross from source to
destination, and has great impact on network performance. As the network size
scales up the probability of processor and link failure also increases. It is
therefore essential to design fault-tolerant routing algorithms that allow
messages to reach their destinations even in the presence of faulty components
(links and processors) . This work focuses on the
development of efficient fault tolerant routing algorithms using probabiltistic concepts for unicast
and broadcast communication on direct networks,
including hypercubic-based and product
interconnection networks.
Networks for Large-Scale Parallel Multimedia Servers
Traditional networks employed in existing parallel computers have been optimised for text data (or best-effort traffic) only. As a
result, they cannot provide guaranteed Quality of Service (QoS)
for the emerging multimedia applications, which combine audio, video and text.
The objective of this research is to design a high-performance interconnection
network suitable for large-scale parallel multimedia servers and evaluate its
performance.
Performance Modelling and Evaluation of Broadcast
Communication
Although many broadcast/multicast algorithms have been proposed for
wormhole-routed networks over the past decade, there has been hardly any study
that proposes an analytical model for these algorithms. This research
investigates the impact of different broadcast algorithms on the
wormhole-routed network under various traffic patterns through analytical modelling. We will study and compare the behaviour of the network using either unicast-based
or multidestination-based broadcast algorithms and
looking for a trade-off between minimisation of
start-up latencies and minimisation of network and
blocking latencies.