With network coding, intermediate nodes between the source and the receivers of an end-to-end communication session are not only capable of relaying and replicating data messages, but also of coding incoming messages to produce coded outgoing ones. It has been the traditional wisdom in information theory that network coding improves the capacity of multicast sessions in directed networks. Studies have also shown that network coding is beneficial for content distribution in peer-to-peer networks, since it resolves the "last block" problem, and eliminates content reconciliation. In this paper, we show that such benefits of network coding does not come without costs and trade-offs. In particular, we refute the previous claim that peers receive linearly independent coded blocks with very high probabilities. Using example scenarios and extensive simulations, we show that it is very likely for peers to receive linearly dependent non-innovative blocks, thus decreasing their efficiency as these redundant blocks consume bandwidth. We observe that such redundancy of network coding is critically dependent on the randomness and sparsity of the P2P topology. We conclude with suggestions on topologies of certain characteristics that are preferred over others, in order to minimize the network coding redundancy, the time to distribute data, and the server cost.

On-demand and live multimedia streaming applications (such as Internet TV) are well known to utilize a significant amount of bandwidth from media streaming servers, especially as the number of participating peers in the streaming session scales up. To scale to higher bit rates of media streams and larger numbers of participating peers, overlay tree or mesh topologies are typically constructed, such that peers utilize their available upload capacities to alleviate the excessive bandwidth demands on stream servers. Previous works rely on random selections of upstream peers, without optimizing the topologies towards maximized utilization of peer upload bandwidth, and as a result, minimized bandwidth costs on streaming servers. We propose Outreach, a distributed algorithm to construct overlay topologies among participating peers in streaming sessions. The design objective of Outreach is to optimize the quality of overlay topologies towards scalability, with respect to the number of participating peers in the session. To be scalable, Outreach seeks to maximize the utilization of available upload bandwidth on each participating peer, and consequently minimize the total bandwidth costs on streaming servers. With analysis, we show that Outreach constructs topologies such that peers can fully utilize their upload capacities, and present a practical distributed algorithm. With simulation-based comparison studies, we show that Outreach effectively achieves its goals in a high-churn peer-to-peer network with an assortment of peer uplink capacities and link delays.

In an intermittent-connectivity network, there rarely exists a connected path between a source node and its destination. These networks arise frequently when each node has a limited transmission range, such as a communication network between separated villages or a surveillance network with a large geographical span. One method of addressing the low connectivity of the network uses redundancy. A node generates and stores data; upon reaching the communication range of another node, it replicates the data to it. Multiple copies of the packet decrease the time to offload the data to the destination, but increase the energy and storage used in the system. In this paper, we quantify the resource-delay trade-off and the throughput capacity for intermittent-connectivity networks with Quality of Service restrictions such as limited communication bandwidth. Many routing protocols have been proposed for these intermittent-connectivity networks. Using the Shared Wireless Infostation Model as an example strategy, we mathematically represent the intermittent-connectivity network and adjust the model to include a Quality of Service constraint. By completely defining a mathematical model, we allow network designers control over system performance through the adjustment of allocated resources such as communication bandwidth, fraction of time a node spends in sleep mode, or required reliability of packet offloading.

Due to their low network connectivity, sparsely-connected networks can support delay-tolerant applications only. Thus, such communication environments are examples of Delay-Tolerant Networks (DTNs). While in some DTN connectivity is predictable (e.g., scheduled), in others, connectivity is created randomly. For instance, to save battery power, nodes in a sensor network may be programmed to follow a pre-arranged sleep patterns, or may enter the sleep state opportunistically. In a sparsely-connected network, source-to-destination route discovery cannot be implemented, since at any particular time, there rarely exists a connected path between the source node and the destination node. Consequently, the conventional approaches to implement network routing do not work and new methods are needed. Flooding (as implemented, for example, using epidemic routing) could be used to spread information through the network. With multiple copies of each packet in the network, the time to offload the data to the destination can be significantly reduced, but at the expense of increased energy and storage utilization. However, in some DTNs, network resources may be limited, for instance due to physical constraints of the network devices. In this paper, we consider the implications of the limited network resources, such as limited communication bandwidth, on the performance and the capacity of a DTN.In particular, leveraging from our previous work on DTN, we use the Shared Wireless Infostation Model (SWIM) to derive a strategy to mathematically represent a DTN, which allows us to evaluate the network throughput capacity, while including a variety of restrictions on network resources. Using this mathematical model derived here, will allow the network designers to adjust the allocated network resources, as to trade off system performance and system resources, while achieving the required network capacity.

It is well-known that live multimedia streaming applications operate more efficiently when organized in peer-to-peer (P2P) topologies, since peer upload capacities are utilized to support other peers, and to alleviate the load and operating costs on the streaming servers. To date, there have been a number of existing experimental proposals with respect to how such peer-to-peer topologies are organized to support live streaming sessions. However, most of the existing proposals resort to intuition and heuristics when it comes to the design of such topology construction (i.e., neighbor selection) protocols. In this paper, we investigate the scaling laws of live P2P multimedia streaming, by quantitatively studying the asymptotic effects and tradeoffs among three key parameters in P2P streaming: server bandwidth cost, the maximum number of peers that can be supported, and the maximum number of streaming hops experienced by a peer. To further generalize our studies, we do not make restrictive assumptions in our theoretical analysis of such scaling laws: both peer upload capacities and peer lifetimes in a session may come from arbitrary distributions. With the theoretical insights we have developed, we propose Affinity, a simple and realistic heuristic to demonstrate the key benefits of our theoretical analysis in dynamic P2P networks, as compared to the topology construction algorithms in existing work.

In wireless ad hoc networks, capacity can be traded for delay. This tradeoff has been the subject of a number of studies, mainly concentrating on the two extremes: either minimizing the delay or maximizing the capacity. However, in between these extremes, there are schemes that allow instantiations of various degrees of this tradeoff. Infostations, which offer geographically intermittent coverage at high speeds, are one such an example. Indeed, through the use of the Infostation networking paradigm, the capacity of a mobile network can be increased at the expense of delay. We propose to further extend the Infostation concept by integrating it with the ad hoc networking technology. We refer to this networking model as the Shared Wireless Infostation Model (SWIM). SWIM allows additional improvement in the capacity-delay tradeoff through a moderate increase in the storage requirements. To demonstrate how SWIM can be applied to solve a practical problem, we use the example of a biological information acquisition system - radio-tagged whales - as nodes in an ad hoc network. We derive an analytical formula for the distribution of end-to-end delays and calculate the storage requirements. We further extend SWIM by allowing multi-tiered operation; which in our biological information acquisition system could be realized through seabirds acting as mobile data collection nodes.