As networking attempts to cleanly separate the control plane and forwarding plane abstractions, it also defines a clear interface between these two layers. An underlying network state is represented as a view to act upon in the control plane. We are interested in studying some fundamental properties of this interface, both in a general framework, and in the specific case of content routing. We try to evaluate the traffic between the two planes based on allowing a minimum level of acceptable distortion in the network state representation in the control plane. We apply our framework to content distribution, and see how we can compute the overhead of maintaining the location of content in the control plane. This is of importance to evaluate resolution-based content discovery in content-oriented network architectures: we identify scenarios where the cost of updating the control plane for content routing overwhelms the benefit of fetching the nearest copy. We also show how to minimize the cost of this overhead when associating costs to peering traffic and to internal traffic for network of caches.

Wireless information-centric networks consider storage as one of the network primitives, and propose to cache data within the network in order to improve latency and reduce bandwidth consumption. We study the throughput capacity and latency in an information-centric network when the data cached in each node has a limited lifetime. The results show that with some fixed request and cache expiration rates, the order of the data access time does not change with network growth, and the maximum throughput order is not changing with the network growth in grid networks and is inversely proportional to the number of nodes in one cell in random networks. Comparing these values with the corresponding throughput and latency with no cache capability (throughput inversely proportional to the network size, and latency of order √n and the inverse of the transmission range in grid and random networks, respectively), we can actually quantify the asymptotic advantage of caching. Moreover, we compare these scaling laws for different content discovery mechanisms and illustrate that not much gain is lost when a simple path search is used.

In this paper, we study the effects of social interactions among nodes on the capacity of wireless networks. We consider three scenarios. In the first scenario, the size of the social group for all nodes is fixed while the frequency of communication within members of a social group follows power law distribution. In the second scenario, scale-free networks are studied where the size of the social group differs from node to node, and the destination in each group is selected uniformly among the members of that group. Further investigation in the second scenario reveals that traditional transport capacity definition provides misleading conclusions for such network models. We show that nodes with different social status impact the capacity differently. By separating nodes with different social status and allocating separate bandwidth to them, it is shown that majority of nodes scale in this network. In the third scenario, both the size of the social groups and the destination in each group are selected according to power law distributions. Our simulation results corroborate the analytical results. Further, we observe consistently that social interaction improves the capacity of wireless networks, which implies that the Gupta-Kumar results were pessimistic for practical networks.

The capacity of a wireless network is studied when nodes communicate with one another in the context of social groups. All the nodes are assumed to have the same number of independent long-range social contacts, one of which each selects randomly as its destination. The Euclidean distance between a source and its social group members follows a power-law distribution and communication between any two nodes takes place only within the physical transmission range resulting in communication over multi-hop paths. The capacity order of such a composite network is derived as a function of the number of nodes, the social-group concentration, and the size of social groups. Our results demonstrate that when each node has constant number of contacts which does not increase with network size growth, and are geographically concentrated, then the network behaves similar to social networks and communication network does not have any effect on the throughput capacity. On the other hand, when the social contact population grows in time, or social connectivity among nodes is highly distributed, then the communication network is the dominant factor and the composite network behaves similar to wireless networks, i.e., the capacity is the same as Gupta and Kumar results. When neither social connectivity nor communication network is dominant, then the throughput capacity results are between these two extreme cases.

We study the impact of social connectivity on the capacity of wireless networks by considering different values of concentration factor and degree dispersion in scale-free networks. The result shows that a capacity similar to Gupta and Kumar [1] is achieved. Further investigation reveals that traditional transport capacity definition provides misleading conclusions for such network models. We show that nodes with different social status impact the capacity differently. By separating nodes with different social status in frequency and allocating separate bandwidth to them, it is shown that majority of nodes scale in this network. The results imply that in a network with social and communication characteristics, social behavior of the nodes has significant influence on the performance of such networks.

We study capacity of social networks when nodes communicate in a wireless environment. Such hybrid networks that are combination of wireless communication and social networks are defined as composite networks. Each node has at least one local contact in each of four directions of the network area and q(n) independent long-range contacts, one of which is selected as the destination. We study the throughput capacity for such networks containing n nodes assuming the same number of social contacts for all nodes. The nodes communicate using multi-hop communications through relaying the packet to one of their local contacts until the packet reaches the destination. The distance between source and its long-range social contacts follows power law distribution with parameter α. The order capacity is derived and compared for different values of α and q(n).

We define composite networks when nodes communicate only with their long-range social contacts and there is no direct link between a node and its long-range contact. Each node has a single long-range contact and all nodes within its transmission range are local contacts for the node. The long-range contact is the destination for each node in the network and since there is no direct link from source to its destination, nodes communicate using multi-hop communications. This is an extension of the famous work by Kleinberg to random wireless ad hoc networks. The throughput capacity of such networks is studied. The routing is based on each node sending the packets to one of its local contacts until the packets reach the destination. The long-range contact distance from a source follows power law distribution with parameter a which is a characteristic of social networks. A tight bound of throughput capacity for different values of a is derived. The results demonstrate that when a increases or equivalently the distance between source and destination decreases, the throughput capacity increases. For α >; 3, throughput capacity of ⊖(1/ log n) is achieved by utilizing simple point-to-point communications where n is the total number of nodes in the network. This is the maximum feasible throughput that can be achieved in point-to-point communications. The result demonstrates the effect of social groups on wireless ad hoc networks. A new parameter called degradation factor is defined which illustrates the asymptotic behavior of networks for large values of n1.