Self-Organization | COMPARING INNOVATION ROUTING

Communication in 4G is expected to live with frequent disconnections due to wide ranging mobility patterns supported including vehicular, user planned disconnection, cellular, short-range and delay tolerant networks. Hence, 4G devices should always be on the lookout to find the best alternative for message delivery depending on availability and application requirements.

Both epidemic and Ant models are leading models for dealing with these challenges by continuously adapting and self organizing. The epidemic approach has proven efficiency for spreading information to all devices with high delivery rates. It nonetheless suffers from some QoS restrictions that could be reduced through buffer management and the use of adequate traffic engineering and routing policies. The probabilistic, random and social metrics can be used to decide whether and when a node must send some information.

On the other hand, the Ant model mimics the search for food, providing a mechanism for finding interesting routing paths according to QoS, security and others requirements. Moreover, the Ant model can increase or decrease the update process of the routing table according to traffic and network stability. When the scenario at hand is unstable, then the algorithm sends many "Forward Ants" to find new paths. Hence, the Ant model may then increase throughput while decreasing delay.

Given that a 4G device may be part of several networks, then it can execute a vertical handoff in order to obtain and exchange information from these different environments, improve its routing and also collaborate with other network users.

In the social routing approach, the nodes can exchange messages in order to identify popular ones and similarity behavior and consequently self-organize to improve their security and performance levels as seen in sections about social routing and social overlay networks.

Routing Optimization | COMPARING INNOVATION ROUTING AND 4G REQUIREMENTS

At the early days of the Internet, routing took into account the number of hops an as important metric for path selection. This was a wise decision at the time as most of the Internet was still homogenous in terms of its links, router capacity and traffic. Soon later, weights were associated to links giving autonomous systems a new criterion to decide on the best routes and a mean to engineer their traffic and balance this. The creation of labels by MPLS provided a similar traffic engineering mechanism capable of controlling and improving the routing and service delivery through path selection.

In the new context of 4G networks, routers must deal with different dynamic link stability levels, security and QoS levels and network handoff It is for such reasons that the 4G networks will certainly need to also consider new routing metrics and change these according to their environment or context. Under some scenarios, reachability could be more important than performance whereas QoS may become the metric of choice in other circumstances. This may also be service and application driven. Electronic mail transfer is a store and forward application that requires information integrity mainly whereas video conferencing considers low delay and bandwidth as primordial network resources.

Future 4G will certainly embrace disruptive connections and delay tolerant networks, high mobility users resulting in a challenging mix with different routing metaphors and techniques thriving within a single 4G unifying architectures. A simple, "one hat fits all" approach to routing cannot be the way forward. Therefore 4G needs to consider multi-metric optimization following different innovative routing approaches instead of merely reusing traditional strategies. It is believed that new routing and resource management insights borrowed from areas as diverse as biology, social phenomena, random and probabilistic diffusion models are expected to lead the way ahead.

But this is nonetheless not a complete breakaway from routing as we know it. In fact one expects to continue making use of useful traditional concepts such as clustering and hierarchical structures to simplify, organize and improve 4G routing. Following a dynamic approach, social algorithms exchange messages to find popular nodes and establish similarity among them in order to create clusters and hierarchical structures. Similarly to traditional routing algorithms from fixed networks, messages can be forwarded from any social node to a popular one judged to be in a better position to disseminate the information and capable of increasing the probability of a message reaching its destination. This offers ways to increase the delivery rate, but differently from the flooding algorithm, the social strategy reduce s the number of message replication as these messages only are forwarding among a restricted number of nodes. Moreover, approaches such as SOLAR and (Leguay, Friedman & Conan, 2006) work by extracting location information to identify mobility patterns in order to improve their routing efficiency. They rely therefore on the understanding of user's behaviors in terms of mobility patterns 4G networks are expected to collaborate with each other independently of their underlying technologies. For example, a user with Bluetooth and GPRS devices can choose one or another technology to disseminate a given type of information according to application level criteria such as urgency and destination distance. One could use an epidemic algorithm to send a simple message to a friend through a Bluetooth interface while selecting a GPRS interface to transfer credit (possible future money) to a distant family member.

Advanced Scheduling Schemes and Resource Allocation

Although the fourth generation is homogeneous under the IP umbrella, the participating 4G devices may have multiple interfaces and radios. A 4G device could bind each one of its interfaces to a distinct network, or use multiple interfaces to access a single network. With these features it will be possible to increase traffic bandwidth using link aggregation or execute hand-off without latency and instability, and maximize processing time available for functions such as data-error correction. Hence future routing protocols cannot assume the presence of static links between device interfaces and networks. They must deal with these heterogeneities under the IP layer and dynamic binding. A new class of challenges emerges, mainly in terms of QoS guarantees. So, how could the user achieve good acceptable performance when using several interfaces submitted to varying working conditions seen at several networks?

Mobile wireless devices often need to maintain data or voice communication across different access points and radio base stations. This process is known as hand-off Current cellular systems implement handoff over a single interface and only for phone calls. However, the next generation wireless system (4G) supports seamless handoff for data traffic and should be able to manage radio resources efficiently. VoIP continuity is another requirement in LTE especially when using the 3GPP IP Multimedia Service (IMS). In such a multi-radio environment, there is space to optimize bandwidth radio resources usage, signal quality and reduce information loss.

An important role for nontraditional routing approaches to play in improving future 4G communication systems is foreseen. The Wasp model has shown to be able to schedule tasks and reallocate resources, following a hierarchical and threshold based approach. Each wasp is stimulated to execute its task when its variable value becomes bellow a given threshold. We have seen this model being applied in the dynamic routing of vehicles, a prominent component of future 4G networks. Here, each vehicle is seen as a wasp with a threshold that waits before finding new optimized paths. When a node receives a request from two or more wasps with the same threshold, it then reserves the necessary resources and improves the routing path to the wasp with the highest hierarchy.

The wasp model may be used to resolve another important problem: that of interface selection. The individual force variation (F) is used for decision making, and to determine the best interface to use. This same wasp characteristic has also been associated to model signal strength, stability, efficiency and power consumption.

Since the binding between network and interface could be seen as a task, then wasp routing (Song, Hu, Tian & Xu, 2005) could also be a good approach to improve routing performance. Moreover, this scheme could also be used to manage and improve robustness by allocating messages to different networks when some paths may become unreachable.