Networks Group
RESEARCH

FIXED AND MOBILE INTEGRATED SERVICES NETWORKS

The field of fixed and mobile integrated services networks, as for example Internet new protocols and architectures, (e.g. IPv6, DiffServ, IntServ), ATM, Gigabit Ethernet, current and future generation mobile and wireless networks, (e.g. Mobile IPv6, WLAN, ad-hoc, GPRS, UMTS, enhanced UMTS) and WDM, is fast moving as a number of emerging (competing) technologies are concurrently reaching stages of maturity in their development. These technologies open up new and exciting opportunities for networking, and are expected to form the cornerstone of the new revolution in the transfer of information in any form, anywhere. High-speed networks integrating voice, image, data and multimedia information are starting to emerge. It is widely accepted by the scientific community that due to the diverse nature of the services that these networks will carry and the network non-homogeneity, network management and control will be very complex, and will require robust, possibly intelligent, control methodologies to obtain satisfactory (if feasible optimal) performance. Toward that goal, the principal aim of our research is to address key issues at a generic level and to apply such theoretical results in the development of efficient and effective management and control techniques. Research issues under consideration include computational intelligence based congestion control, non-linear and adaptive control based congestion control, control structures and techniques for effective control, network survivability, and optimisation of resource allocation using computational intelligence. Much of this work is conducted in collaboration with Ahmet Sekercioglu (Centre for Telecommunication and Information Engineering, Monash University, Melbourne, Australia), Petros Ioannou (University of Southern California), Hugo de Graaf , Vlora Rexhepi, Anthony Lo (Ericsson Eurolab Netherlands, ELN), Geert Heijenk (University of Twente, Netherlands), Manuel Dinis, Jaime Ferreira (Portugal Telecom Inovacao, PTI), Americo Correia, Fernando Velez (ADETTI-Portugal), Antonio Rodrigues (Technical University of Lisbon, Portugal), Athanasios Vasilakos (FORTH, Crete, Greece), G. Ramamurthy (NEC Research Labs, USA), Costas Pattichis, George Samaras, Marios Dikaiakos (University of Cyprus), and Soulla Louca (Intercollege, Cyprus).

Top

Wireless and Mobile Networking

A. Pitsillides, C. Chrysostomou, N. Jacovides, J. Antoniou, G. Hadjipollas, S. Tsakkiouris (University of Cyprus), Hugo de Graaf , Vlora Rexhepi, Anthony Lo (Ericsson Eurolab Netherlands, ELN), Geert Heijenk (University of Twente, Netherlands), Manuel Dinis, Jaime Ferreira (Portugal Telecom Inovacao, PTI), Americo Correia, Fernando Velez (ADETTI-Portugal), Antonio Rodrigues (Technical University of Lisbon, Portugal).

This past decade, we have witnessed the tremendous growth of the Internet and the huge success of second generation (2G) digital wireless standards. The research community is now directing its interest towards unified ways of looking at system design, optimization, and Quality of Service (QoS) issues to satisfy the requirements of next generation mobile networks. These developments have also been the main drivers for Internet Protocol (IP) based mobile networks.
The implementation of IP-based transport networks in future wireless networks implies that IP QoS architectures and mechanisms will be used. Considerable work has been done in both, the development of a framework for Internet QoS and the design of IP-based wireless network architectures. The knowledge gathered provides a good foundation for the development of resource management and congestion control mechanisms for third generation (3G) Universal Mobile Telecommunications System (UMTS) networks and beyond.
Currently, the existing resource management and congestion control mechanisms are not able to cope with the requirements implied by IP-based transport networks as envisaged in future wireless network architectures.
As part of our research, we have proposed a generic Integrated Dynamic Congestion Control (IDCC) scheme using non-linear control theory for congestion control within the Differentiated Services (DiffServ) framework, and a hierarchical fuzzy logic control architecture aimed at handling the complexity of the IP-based UMTS network.
The rapid growth of the Internet and increased demand to use the Internet for time- sensitive applications necessitate the design and utilization of new Internet architectures, like the DiffServ architecture, with effective congestion control algorithms like the IDCC scheme. Network congestion control remains a critical and high priority issue, even for the present Internet architecture. Furthermore, congestion may become unmanageable unless effective, robust, and efficient methods for congestion control are developed. Consequently, the need for providing QoS support and tightly controlling traffic in a UMTS IP-based RAN (Radio Access Network) motivates the formulation of appropriate control strategies in the same spirit as IP DiffServ.
As part of our involvement in the IST-funded SEACORN project on Enhanced UMTS (E-UMTS) networks we aim, among other things, to: (i) develop and implement resource management algorithms enabling QoS provisioning and differentiation while optimising resource efficiency in a multi-service, multi-rate UMTS network, and (ii) develop and implement resource management algorithms for IP-based core networks. Our objective is to use IP as an end-to-end networking solution for E-UMTS networks. This research will focus on IP version 6 (IPv6) since it is widely considered as the most flexible networking protocol currently available.

Top

Congestion Control

A. Pitsillides, C. Chrysostomou, L. Rossides (University of Cyprus), Ahmet Sekercioglu (Centre for Telecommunication and Information Engineering, Monash University, Melbourne, Australia ), P. Ioannou, M. Lestas (University of Southern California), G. Ramamurthy (NEC Research Labs, USA).

It is generally accepted that the problem of network congestion control remains a critical issue and a high priority, especially given the growing size, increasing demand for new services with varying quality of service characteristics, and higher speed (bandwidth) demanded from an increasingly integrated services network. One could argue that network congestion is a problem unlikely to disappear in the near future. Furthermore congestion may become unmanageable unless effective, robust, and efficient methods for congestion control are developed. The existing congestion control solutions deployed in the Internet Transport Control Protocol (TCP), based on the so called end-to-end approach, for TCP transported traffic are increasingly becoming ineffective, and it is generally accepted that these solutions cannot easily scale up even with various proposed "fixes". In this approach sources implicitly infer congestion from lost packets, and in such event they reduce their rate substantially. This approach served the Internet well and was appropriate for pure best-effort data carried by TCP, with little or no sensitivity to delay or loss of individual packets. Active Queue Management (AQM) was proposed, which potentially offers better performance. In AQM, routers can detect congestion before the queue overflows, i.e. congestion feedback is no longer limited to packet drops as an indication of congestion. Random Early Discard was a leading example, demonstrating better performance, and currently reaching implementation in commercial routers. However, performance problems were shown, due to the ad-hoc design approach and the subsequent difficulty of tuning the parameters (which are dependant on network conditions and are sensitive to number of active flows) and its reliance on the existing TCP congestion control framework. Feedback (implicit) in this case was provided early due to the active policy of dropping packets, dependant on the level of the queue length. Recently, Explicit Congestion Notification (ECN) was adopted by the IETF as an Internet standards track protocol. In ECN, active queue management routers are allowed to use the Congestion Experienced (CE) code point in a packet header as an indication of congestion, instead of relying solely on packet drops. This has the potential of reducing the impact of loss on latency-sensitive flows, as well as reducing excessive delays caused by retransmissions after packet drops. It is also worth noting, that for Asynchronous Transfer Mode (ATM) a similar approach was witnessed, with the performance of the vast majority of the congestion control schemes proposed not proven analytically.
Most proposed schemes are developed using intuition and simple (ad-hoc) non-linear control designs. These have been demonstrated to be robust in a variety of scenarios that have been simulated. Since these schemes are designed with significant non-linearities (e.g. two-phase-slow start and congestion avoidance-dynamic windows, binary feedback, additive-increase multiplicative-decrease flow control etc), the analysis of the closed loop behavior is difficult if at all possible, even for single control loop networks. The interaction of additional non-linear feedback loops can produce unexpected and erratic behavior. Empirical and analytical evidence demonstrates the poor performance and cyclic behavior of the controlled TCP/IP Internet. This is exacerbated as link speed increases to satisfy demand (hence bandwidth-delay product, and thus feedback delay, increases), and also as the demand for better quality of service increases. Clearly, more effective congestion control schemes are needed to prevent serious economic losses and possible "meltdown" of the Internet.
In our work we adopt different techniques for solving the congestion control problem:

Non-linear Adaptive Congestion Control

Despite the successful application of control theory to other complex systems the development of network congestion control based on control theoretic concepts is quite unexplored. Most of the current congestion control methods are based on intuition and ad hoc control techniques together with extensive simulations to demonstrate their performance. The problem with this approach is that very little is known why these methods work and very little explanation can be given when they fail.
Recently several attempts have been made to develop congestion controllers, mostly using linear control theory. Despite these efforts the design of congestion controllers whose performance can be analytically established and demonstrated in practice is still a challenging unresolved problem. Lately a serious attempt was made to model TCP/AQM and to use control theory to address the congestion control problem. The proposed PI controller for AQM uses classical control system techniques to design a control law for the router queue management. The non-linear dynamic model for TCP/AQM is linearized around an operating point, and a stable PI linear controller is designed. The derived controller suffers the disadvantage that it is unable to maintain its performances as the network state changes (moves away from the assumed operating point), as for example when the number of TCP flows increases. This in effect produced sensitivity and stability problems. For example, tuning based on a small number of flows can lead to stability problems when the actual number of flows is large, whereas tuning based on a high number of flows can lead to sluggish responses, when the actual number of flows is underestimated. However, we believe that the richness of non-linear control theory developed during the recent years justifies its use now. Our preliminary results are promising. Using non-linear control theory, a generic Integrated Dynamic Congestion Control (IDCC) scheme is recently proposed to implement a new service model where applications with different quality of service requirements are accommodated in different classes of service so as to maximize the overall utility function in the fashion of Shenker (1995). The IDCC scheme is based on a non-linear model of the network that is generated using fluid flow considerations. Currently, we proceed to further studies on the implementation of IDCC within a Enhanced UMTS environment.

Top

Fuzzy Logic based Congestion Control

Designing effective congestion control strategies is known to be difficult because of the complexity of the structure of the networks, the nature of the services supported, and the variety of the dynamic parameters involved. Consequently, a number of researchers are also looking at alternative, non-analytical, control-system design and modelling schemes that have the ability to cope with these difficulties in order to devise effective, robust congestion control techniques to supplement, or even replace, traditional control approaches. These schemes employ artificial neural networks, fuzzy systems, and design methods based on evolutionary computation (collectively known as Computational Intelligence). Currently, the application of fuzzy control techniques to the problem of congestion control in IP-based networks is suitable due to the difficulties in obtaining a precise mathematical model using conventional analytical methods. Moreover, traffic congestion on the Internet is a concept which is well understood. Therefore it is possible to obtain simple linguistic rules for congestion control. Such a fuzzy logic-based queue management is already proposed. Our design of a fuzzy control system for DiffServ networks is based on a fuzzy logic controlled Random Early Detection (RED) queue, which can provide good results in the presence of dynamic network state changes. More precisely, a Fuzzy Inference Engine (FIE) is designed which uses separate linguistic rules for each predefined class in the router queues to preferably drop packets in DiffServ networks. The proposed fuzzy logic strategy is shown to be robust with respect to traffic modelling uncertainties and system nonlinearities, yet provide tight control. As a result, it offers good service. Furthermore, a new active queue management scheme, Fuzzy Explicit Marking (FEM) is recently proposed, implemented within the Differentiated Services framework to provide congestion control using a fuzzy logic control approach. The newly proposed scheme allows the use of linguistic knowledge to understand the dynamics of nonlinear probability marking functions, and so more effective implementation, use multiple inputs to capture the dynamic state of the network more accurately, achieve finer tuning for packet marking behaviours (either dropping the packet or setting its ECN bit) for aggregated flows, and thus to provide better QoS to different types of data streams, whilst maintaining high utilization. Further enhancements to FEM scheme are under study, such as adaptation techniques, Call Admission Control techniques, introduction of a fuzzy logic-based supervisory mechanism for coordinating the various proposed algorithms within an IP-based Diff-Serv environment (both fixed and mobile networks).

Top

Hierarchical Fuzzy Logic Control Structure

There exists a large number of published works solving isolated (individual) control functions, such as congestion control, connection admission control, resource management, and network survivability in large scale telecommunication networks. It would be desirable to undertake a thorough study of an integrated control structure implementing a multi-level control strategy spanning the network, call, and packet levels. The integration problem is extremely complex. So far, the proposed schemes have been motivated by the implementation of one solution to the problem, organized in a multi-level structure. It seems likely that heuristic methods employing fuzzy logic formulations will help to harness human visualization and reasoning skills in the integration process, and so fuzzy logic approaches are likely to ease the integration problem. The integration of the control functions can be achieved by appropriate reformulation of the existing and separately designed strategies in a new multilevel fuzzy logic structure, and/or conceiving new ones applied in an IP-based UMTS RAN environment, with their coordination achieved via a hierarchy of fuzzy logic based supervisors.

Top

Adaptive Fuzzy Logic Systems

Even though fuzzy control may be designed from information based on easily understood linguistic knowledge, in order to achieve optimum performance, some of the selected parameters must be tuned manually, usually by observing the behaviour of the controlled system. To automate the tuning procedure some form of adaptability needs to be provided. The main emphasis has been in the formulation of an adaptive fuzzy controller that is synthesised from a collection of fuzzy IF-THEN rules. These fuzzy rules are either collected from experienced human operators, generated automatically during the adaptation procedure, or a combination of the above. Our current work assumes that a collection of IF-THEN fuzzy rules has been selected and provides adaptation by adjusting certain parameters of the membership functions characterising the linguistic terms in the fuzzy IF-THEN rules, in order to control a system to a desired state. The behaviour, in terms of the convergence of the adaptive algorithm and global stability and robustness of the overall adaptive fuzzy logic system (for different adaptation strategies), is currently investigated using analysis and simulation. It is expected that the adaptive fuzzy controller will be useful in applications where a large degree of uncertainty about the system and its behaviour is present, such as in wide-area high-speed networks.

Top

Predictive Adaptive Congestion Control

The large bandwidth-delay product (large propagation delay in comparison to the buffer dynamics) and the uncertainty in traffic modelling make the design of congestion controls difficult. To address these problems we investigate the suitability of adaptive predictive control techniques, featuring both feedback and feedforward, in the solution of the combined connection admission and flow control problem. We provide effective prediction, beyond the propagation delay, by focusing on the net input traffic (rather than the buffer length) whose time dynamics, due to the high correlation present in some Variable Bit Rate (VBR) traffic, are comparable to large propagation delays. Using analysis and simulation we show that high utilisation of resources can be achieved while maintaining guaranteed quality of service to the user. Also the proposed scheme displays adaptability, robustness, and tolerance to fairly long propagation delays, and random connections and disconnection of sources.

Top

Intelligent Optimization Techniques

A. Pitsillides, C. Pattichis (University of Cyprus), A. Sekercioglu (Centre for Telecommunication and Information Engineering, Monash University, Melbourne, Australia), and A. Vasilakos (FORTH Crete)

Network optimisation problems, such as bandwidth allocation to Virtual Paths (VPs), generally belong to the class of multi-objective non-linear constrained optimisation problems. Currently, we concentrate our efforts on intelligent optimisation techniques, such as techniques based on genetic algorithms (GAs), evolutionary strategies and Evolutionary Programming (EP). GENOCOP, a GA algorithm for Numerical Optimisation for Constraint Problems was applied in the problem of VP bandwidth allocation) in ATM based networks, initially for the single optimisation case. GAs facilitate an efficient non-linear function optimisation paradigm. Our preliminary findings suggest that GAs offer a satisfactory solution to the problem of bandwidth allocation for virtual paths, exhibiting a very fast convergence and fairness in the allocation of bandwidth to VPs. Further work is currently in progress to explore completely the capacity of GAs for more complex topologies, not only for single objective optimisation, but also for the multi-objective case. For Internet based networks, current trends are emerging to allocate bandwidth to different classes and users in order to support QoS provisioning. This despite the complexity introduced when compared with the current Internet Best-Effort model. The architecture of various proposed solutions differ in detail, but the underlying model is similar in the sense that network resources need to be allocated and policed. Aggregated bandwidth allocation, without the need for per-session signalling, is currently advocated for DiffServ Internet architectures.

Top

Wireless Computing and e-Services and Telehealth Care

A. Pitsillides, G. Samaras, M. Dikaiakos (University of Cyprus), E. Christodoulou (NetU company, Cyprus)

Mobile networking is an enabling technology for the new paradigm in nomadic computing. In this research we aim to and focus on wired and wireless networks that facilitate the transfer of information, their management and control, and co-ordination software that ties the computing elements together, and e-Services. In earlier work, Computational Models for Wireless Computing, we proposed a new Client/Intercept Computational Model for Wireless Environments. We extend this work in support of Virtual Collaborative Teams in Telehome Health Care Systems. DITIS is a two year funded research program, which is focused on the home treatment of cancer patients in Cyprus. A pilot has already been underway in Larnaca. Currently phase II is being implemented which will extend DITIS to the whole of Cyprus as well as link the Bank of Cyprus Oncology Centre with the DITIS project for home care.

Top

Network Survivability using Graph Theoretic Techniques

A. Pitsillides, G. Samaras (University of Cyprus), S. Louca (Intercollege, Cyprus)

Due to the increasing reliance of our information hungry society on the timely and reliable transfer of large quantities of information such as voice, data, and video across high speed communication networks, it is becoming important for a network to offer survivability, or at least graceful degradation, in the event of network failure. Survivability techniques can be classified as network design and management procedures to minimise the impact of failures on the network. These can be grouped into three categories: prevention, network design, traffic management, and restoration. The objectives of this work are to formulate and develop generic algorithms which can be applied to improve network survivability. Our initial work concentrates on the multi-path routing problem. We use graph theory to select the K-best disjoint paths, between any origin destination pair, which can be used for load balancing and rerouting. We provide a number of algorithms that transform a given network into a trellis graph, and consequently use it with an algorithm that transforms the trellis graph into an equivalent minimum cost network flow problem to find the K-best path through the trellis. The significance of this work will be its fundamental contributions to the base knowledge in network survivability techniques.
In a recent proposal, given any random topology, network survivability can be assisted by finding in polynomial time disjoint paths to allow alternative routing.

Top

Intelligent QoS Routing

A. Pitsillides (University of Cyprus), A. Vasilakos, K. Anagnostakis, (FORTH Crete, Greece), C. Ricudis (independent researcher), W. Pedrycz (University of Manitoba, Canada)

One major issue in future high-speed communication networks is the problem of routing, given QoS and policy constraints. The problem has been proved to be extremely complex and heuristics have been proposed to reach near-optimal solutions. In this study we propose the use of computational intelligence algorithms, such as fuzzy logic and genetic algorithms for route computation. We designed an inter-domain routing mechanism for global broadband connections that uses a fuzzy-logic controller to compute route suitability. Our work resulted in a system that is efficient, self-improving through use of genetic algorithms, and can be easily adapted to other environments as well.

Top

Parallel Simulation of Large-scale Telecommunication Networks

A. Pitsillides, J. Antoniou (University of Cyprus), A. Sekercioglou, G. Egan, A. Varga (Centre for Telecommunication and Information Engineering, Monash University, Melbourne, Australia)

Parallelization, as a concept, is the idea of distributing the work that could take up a lot of resources and time on one processor to many processors. This is achieved by giving a small task to each processor and combining the different outputs, therefore, achieving significantly faster results. Furthermore, by distributing a network model on several machines, the memory requirements on any single system are decreased. This is important because large network simulations usually need excessive memory as well as CPU time.
The main objective of this research is to build a high-performance discrete event simulation framework for performance analysis of large-scale broadband telecommunication networks. We plan to use the framework to analyze the performance of protocols for provision of QoS and predictive congestion control mechanisms in large-scale telecommunication networks via simulation studies. We specifically intend to study delivery of multimedia through the wired and wireless networks.
Through our involvement in the IST-funded SEACORN project we are developing a system-level real-time simulator for an Enhanced UMTS network. A high-performance distributed platform that supports parallelization will be used for the simulation experiments.

Top