Abundant networks and services
The SRA theme Computer Networks addresses the issue of how to offer anybody, anytime, anywhere, “liquid bandwidth”, through a heterogeneous network of (local) wireless links and (global) broadband fiber networks. Key aspects are adaptively, dependability, interoperability, and maintainability The abundant and increasing availability of wireless connectivity, (e.g., in consumer-installed wifi access points and operator-owned cellular networks), and the potential of using wireless multihop techniques, such as in mesh or ad-hoc networks, provides the potential for omnipresent, abundant network access for users, services and devices. For a specific user, service or device, it is highly likely that at a certain location and time, communication means will be available for him to communicate through a broadband connection with other users, and services and devices somewhere in the world.
Aim and Mission
The SRA theme Computer Networks addresses the issue of how to incorporate all these (wireless) links into a system providing “liquid bandwidth”, anytime anywhere, like IP and web technology have done for computer networks. Prerequisites for such a system are the availability of ubiquitous computing, wireless communication, and broadband networks. The main challenge is how to bundle all communication and computation (“compunication”) capabilities that are often “owned” by users themselves (e.g. wifi access points), into a cooperative, self-organizing, adaptive grid, where each user is both producer / provider and consumer / user of (communication / content) services. Such cooperative networking should ensure that abundantly available, uncoordinated, and unpredictable “compunication” facilities can support processes in our society.
Examples of applications are to further increase safety and efficiency in transport, e.g., by using car-to-car communication, to enhance personal comfort and safety by using personalized networks, to enhance organizational efficiency by using trans-sectoral networks and composable business support services, and to support processes in specific domains like healthcare for cure, care and wellbeing through continuous and unobtrusive health monitoring.
Future heterogeneous networks need to possess high degrees of autonomy and self configurability because classical operation and maintenance methods are hardly applicable to these network types. We envisage the heterogeneous network of the future to consist of some ad-hoc elements combined with infrastructure based network support. Communication over short distances and with users in the direct vicinity could be supported in ad-hoc mode, while long-distance communication would make use of the existing fixed infrastructure. In this network, peer-to-peer communication between any two nodes could be supported, generating an overlay network structure.
Current and future mobile and wireless networks enable new services for end-users/consumers that should be user-centric (e.g. I-centric or We-centric). Working from the premise that users are connected virtually at any time and any place, new ways of interacting among users and between users and networked services become possible.
In the future new business models may prevail overcoming the traditional hierarchy of consumer, service provider and network provider. The traditional distinction between a user and provider will blur. Typically, the end-user that now predominantly consumes services will change into a provider of communication and networked services. Blogging is possibly an early example of an end-user provided information service. User-centric computing and networking may also lead to the extension of the peer-2-peer model for sharing content to a cross-cutting peer-2-peer model for sharing communication, computation and information resources. Applying a service-oriented approach to user-centric computing and networking is a promising way forward. Challenges that need further research include the mechanisms to publish services (i.e., service directories), service discovery (i.e. interworking between heterogeneous service discovery protocols) and service binding (i.e. session management and control in case of a disruptive networking environment).
Design of heterogeneous networks
To support the design of these heterogeneous computer networks generic methods, techniques and tools are needed to assess the performance, dependability and security (and possibly evolvability and maintainability) of these networks, i.e. performance expressed in terms of throughput, delay or loss measures, given a certain traffic mix, and dependability in terms of minimum availability over each year, or the maximum time between outages, and so on. To be able to analyze designs, the design itself has to be described in a formalized language, not only to proof functional correctness, but especially also to be able to verify whether performance and dependability characteristics are met. Using techniques known from stochastic operations research (queuing theory, stochastic processes, Markov chains, discrete-event simulation), these formalized designs should be analyzed to certify adequate performance and dependability. At the same time, preferably from the same formalized design, functional properties should be verified, using techniques like model checking. Next to that, graph theoretical methods are needed to assess topological issues, especially in large IP-based networks; such topological descriptions can be used to characterize traffic streams, hence, form important input to the stochastic models that are used to obtain performance characteristics. Next to model-based analysis methods to support the design, in later phases of the design process, when parts of the system are operational, measurement techniques will play a more important role. Such techniques require a detailed knowledge and insight in operational networking systems and their protocols, as well as in statistical techniques to capture the necessary information in a correct (unbiased) and non-intrusive way. Measurements also form the necessary input to workload characterization methods, in which measured workloads (traces) are used to parameterize easy-tohandle stochastic workload models.
Self-organization and self-management
A key aspect in future networking systems will be their ability to adapt at runtime, e.g., based on context-awareness or any other self-awareness (including self-reflection; cf. reflective systems). Aim of adaptation is to self-tune the system’s functionality (services) and performance and/or dependability automatically and autonomously. When systems increase in size and complexity, this is the only viable approach to pursue, since one cannot anticipate all possible system and environmental configurations a priori. A promising technique in this respect is model-based adaptation, in which the adaptation process is governed by an underlying mathematical system model. Notice that in this approach, the adaptation process in fact comprises a very general class of closed-loop control systems; adequate control theory for this type of systems is currently not well developed. Although many different technology-based solutions exist for adaptive systems, an overall architecture in which the different technologies can be positioned and related to each other, is still not available.
Given that we expect every artifact to be equipped with computing and communication capabilities, we have to face a large heterogeneity of device capabilities and energy supply. Already now we see systems consisting of powerful computers on the one hand and large numbers of cheap disposable sensor nodes (e.g. RFID tags) on the other hand. The range of processing power, storage capacity, communication interfaces and functionality will be very wide. Moreover various wired and increasingly wireless link technologies with very different characteristics will coexist. The way networks are managed nowadays by specialized system managers is not scalable to these future networked embedded systems, because of the sheer numbers of components, the system dynamics, the heterogeneity of devices and subsystems. Systems need to be self-configurable and selfoptimizable.
Self-configuration and reconfiguration implies that components are able to bootstrap without operator intervention, detect their environment (neighboring components) and context, are able to connect to other components and organize themselves in a network. This process may, depending on the complexity and the extent of the physical distribution of the system, have to repeat itself at a higher level and form overlay networks between components and subnetworks. The system should be conscious of its purpose, i.e., the applications it has to support and form and configure itself accordingly.
Self-optimization is a requirement, since the resources in future wireless ad-hoc networks are likely to be limited (e.g., energy supplies, capacity of radio channels), and, operating conditions and system mission may change in unpredictable ways. Ad-hoc networks should be able to capture and act upon changes in their environment, e.g., car networks may want to sense the presence of other networks belonging to other cars and, depending on the speed of the car decide to link up and start up cooperative applications to enhance safety. This behavior is called context aware. Context awareness is a very active research theme; however context in relation to driving the formation and self-organization of embedded networks is only starting to be addressed.
Finally, networks should be reliable. People will start to rely on them heavily, without noticing their presence. They will form the backbone of automated systems, e.g., for vehicle safety and protection against industrial accidents. They will also have to be trustworthy, i.e., they will have to satisfy stringent requirements regarding dependability, security and, for person oriented applications, privacy. This implies the aforementioned self-healing and reconfiguration, but also self-protection against threats, because of the limits on the response time to threats.
Application areas
Personal networks, Emergency networks, Sensor networks, Home networks, Access networks, Ad hoc networks, Vehicle networks, Health monitoring, Economic impact can be expected on the ICT sector, transport sector, health sector, public sector and safety sector.
