Broadband Communications Networks

 

 

To realise the application aware network services envisaged for next generation networks, all layers need to communicate and exchange the required information. However, signalling needs to re-main scalable and thus the less is required the better it will perform. Today IP serves as hard informa-tion border, and this is one of its primary features. What is above does not need to care for what is below and vice versa. Thereby technologies and protocols above and below became independently exchangeable. This flexibility needs to be preserved. Still, if the lower layers have no idea of the demands the higher layers rise, they cannot support them efficiently. On the other side, if the upper layers ignore the conceptual restrictions of lower layers, the services they offer cannot work properly.

Research Topics

Research Areas

To realise the application aware network services envisaged for next generation networks, all layers need to communicate and exchange the required information. However, signalling needs to re-main scalable and thus the less is required the better it will perform. Today IP serves as hard informa-tion border, and this is one of its primary features. What is above does not need to care for what is below and vice versa. Thereby technologies and protocols above and below became independently exchangeable. This flexibility needs to be preserved. Still, if the lower layers have no idea of the demands the higher layers rise, they cannot support them efficiently. On the other side, if the upper layers ignore the conceptual restrictions of lower layers, the services they offer cannot work properly.

Good planning of the capacity distribution and the network operation schemes, algorithms, and protocols is mandatory. The tasks related to this are summarised by the term network design. New methods and approaches may be developed and applied layer per layer. Particular attention should be paid on improving performance, reliability and energy efficiency of entire network infrastructures. Also implementation issues need to be considered, especially migration options from an existing architecture towards an improved one. Once a network is realised, it needs to be managed. The tasks related to optimising or at least maintaining the network’s performance at different loads are sum-marised by the term network engineering.

Modern operator networks are commonly composed engaging several switching layers. For example a legacy IP over Ethernet over SDH infrastructure encompasses three switching layers: packet switching (IP), frame switching (Ethernet) and circuit switching (SDH). Cross-layer design and optimisation are challenged heavily by the correlations and dependencies that these parallel network layers cause. The legacy approach is to design (semi-)static virtual topologies in the lower network layers to serve the upper. However, the design of these is a complex issue that demands to balance efficiency against complexity, especially if the demands change on a short time scale and the solutions average life-time is short. Generalised multi protocol label switching (GMPLS) hierarchically combines the management in a vertical control structure, whereas alternative concepts (e.g., flow transfer mode – FTM) intend to achieve a horizontal integration of the control issues.

Network design and optimisation (in total and layer by layer) are two important research areas of the Networks Group. To perform both effectively and efficiently, abstract modelling of network protocols, topologies and architectures (i.e., implementations) are essential. Selecting fitting, still solveable, models is a key challenge, besides solving, analysing and thoughtful interpreting of gained results. Another important research topic is analysing and evaluating proposed improvements, i.e., monitor-ing and contributing to the development of common standards of the IT industry in combination with performing tests in the laboratory and exemplary monitoring of network performance in the field.

Network Mathematics

Queueing models are commonly analysed using state-transition diagrams. These are very convenient to study dynamic (sub-) systems in an encapsulated fashion. The system of equations to be solved results from the equilibrium equations and the condition that the system needs to be in one state at any time. Several performance metrics can be directly calculated from state probabilities, e.g., system utilization and blocking probability. With Little’s formula the flow- and waiting time can be calculated from system- and queue filling, respectively. However, if the number of states is huge solving can become mathematically intractable. To achieve approximate results a smaller system may be extrapolated or a similar infinite system truncated.

The derivation of distributions via Laplace transforms is mathematically challenging but the most powerful method to calculate traffic characteristics that result from combining of processes. The formulation of stochastic and performance related problems in the transform domain enables closed-form solutions if inversion is traceable. Recent mathematical publications present new methods and open a new space for practical engineering using mathematical analysis. Sometimes solving demands coarse simplification, which queries the analysis precision. Less complex approximate methods typically simplify the solving issues. However, their merits and limitations need to be known well in order to not misinterpret the results.

Network Technologies

Basic knowledge about key technologies and methods for the implementation of different network functions as well as possible realizations of network elements and systems are today broadly required to understand the features, limits, and developments of the current communication networks.

The topic reaches from hardware components, systems and interfaces, including electronic as well as all-optical and opto-electronic variants, to network element designs (structures and topologies inside the network elements, i.e., switch architectures, routing protocols, and network control units and interfaces), and network control and management mechanisms and strategies, all the way to the applications and services that these building blocks enable.

Global networks are usually composed engaging mixtures of technologies and protocols. Additional to the implementation of network elements employing photonic and electronic technologies and using wired or wireless transmission, various protocols and signalling methods as well as routing algorithms are required. For a future-prove and sustainable global communication network are energy efficiency, reliability and security as crucial as advanced, demand specifically adjustable (application aware) communication services. In this spirit, an optimal support of new applications such as various web applications, high-definition and interactive television, videoconferencing, building energy management, e-mobility and smart metering is an important goal.

Technologies covered include FTTX, WDM, OBS, OPS, OTN, SONET/SDH, Ethernet, FC, ATM, xDSL, IPv4, IPv6, MPLS, LAN, WLAN, WiFi, WRAN, GSM/GPRS, UMTS, LTE, RoF, CATV, SAN, satellite communication, network interfaces, and networked interconnects.

The conversion of formerly independent specialist data-networks into a heterogeneous interconnected public infrastructure (the Internet), accessible from a wide variety of people and devices, causes sever privacy and security issues. New protocols and encapsulation schemes have been added, but to be successful these may not degrade the established simplicity of using network services. Therefore, they need to be autonomous and invisible to customers, and at the same time they need to ascertain the customer that they are in operation and assure the privacy of all communication at all times. Evidently, proposed security and privacy schemes demand to be reliably evaluated and tested prior implementation. However, the plurality of technologies and legacy protocols that may influence them makes this a laborious mission. The necessity for secure, reliable and at any time available access to the Internet increases by another magnitude, if we consider the envisioned next transformation of the Internet into a cloud that not only offers information and services, but also storage and processing (cloud computing) capabilities. In parallel to the increasing importance of network services to people, also the support of mobile services increases, i.e., the demand to provide access to the same service from any place using any equipment at any time.

Network Performance

The Internet Protocol (IP) was designed to provide robust connectivity whereas not guaranteeing delivery times of packets. Caring for losses was even left to the applications, and latency was not a severe issue. Today, data networks are over-dimensioned to provide a reasonable service quality. Introduction of time sensitive applications has suddenly launched challenges regarding service quality. Increasing the over-provisioning ratio to levels that fit every possible future demand is economically impossible. The modern approach is service differentiation. To realise this within a single network layer we need to reconsider common practice, and validate its mechanisms for any network condition.

In data networks, temporary load peaks exceed the average load considerably. Consequently, we observe an exponential decrease in performance when the average offered load approaches the provided capacity. Therefore, a properly planned data network provides considerably more resources than average load would suggest. Load limiting, policing, and shaping are commonly applied to reduce the congestion potential and thereby increase the transport performance and average quality of service (QoS). Prioritization introduced to privilege latency sensitive services and to implement QoS differentiation, and label switched paths (LSPs) as well as other concepts are use for load balancing, i.e., to put the traffic where the resources are and determine the handling of individual traffic flows at intermediate nodes. Continuous monitoring and managing of both traffic and network resources can be used to better and faster react on changing conditions, which can lead to an improved performance and a better utilisation of available resources.