Given the increasing popularity of real-time video communication over enterprise wireless networks, ensuring good quality of experience is becoming critical. A common problem in such networks is that the clients typically adopt the strategy of associating to the access point with the highest signal strength, which can lead to load imbalance, unfair throughput allocation, and low overall video quality. In this paper, we propose to improve load balancing across video clients by selectively triggering directed roams of chosen clients, while explicitly accounting for the cost of roaming. We formulate the problem of identifying the directed roams to perform as an optimization problem where the objective function reflects both the overall utility of all video clients and the cost due to roaming. We subsequently apply relaxations to the original NP-Hard problem, and propose a three-stage algorithm that efficiently computes an approximate solution. Simulations show that by selectively roaming just 5% of the clients, we increase throughput relative to an initial strongest signal based association by at least 10% for over half of the clients and by at least 20% for over a quarter of the clients. We also increase aggregate throughput, improve fairness and thereby increase the overall video quality.

Reliable packet delivery within stringent delay constraints is of paramount importance to mission-critical computer applications with hard real-time constraints. Because retransmission and coding techniques counteract the delay requirements, reliability may be achieved through replication over multiple fail-independent paths. The existing solutions, such as the parallel redundancy protocol (PRP), replicate all packets at the media access control layer over parallel paths. PRP works best in local area networks; however, it is not viable for IP networks that are a key element of emerging mission-critical systems. This limitation, coupled with diagnostic inability and lack of security, renders PRP unsuitable for reliable data delivery in these IP networks. To address this issue, we present a transport-layer solution: the IP parallel redundancy protocol (iPRP). Designing iPRP poses nontrivial challenges in the form of selective packet-replication, and soft-state and multicast support. iPRP replicates only time-critical unicast or multicast user datagram protocol traffic. iPRP requires no modifications to the existing monitoring application, end-device operating system, or to the intermediate network devices. It only requires a simple software installation on the end devices. iPRP has a set of diagnostic tools for network debugging. With our implementation of iPRP in Linux, we show that iPRP supports multiple flows with minimal processing-and-delay overhead. It is being installed in our campus smart-grid network and is publicly available.

Reliable packet delivery within stringent delay constraints is of primal importance to industrial processes with hard real-time constraints, such as electrical grid monitoring. Because retransmission and coding techniques counteract the delay requirements, reliability is achieved through replication over multiple fail-independent paths. Existing solutions such as parallel redundancy protocol (PRP) replicate all packets at the MAC layer over parallel paths. PRP works best in local area networks, e.g., sub-station networks. However, it is not viable for IP layer wide area networks which are a part of emerging smart grids. Such a limitation on scalability, coupled with lack of security, and diagnostic inability, renders it unsuitable for reliable data delivery in smart grids. To address this issue, we present a transport-layer design: IP parallel redundancy protocol (iPRP). Designing iPRP poses non-trivial challenges in the form of selective packet replication, soft-state and multicast support. Besides unicast, iPRP supports multicast, which is widely using in smart grid networks. It duplicates only time-critical UDP traffic. iPRP only requires a simple software installation on the end-devices. No other modification to the existing monitoring application, end-device operating system or intermediate network devices is needed. iPRP has a set of diagnostic tools for network debugging. With our implementation of iPRP in Linux, we show that iPRP supports multiple flows with minimal processing and delay overhead. It is being installed in our campus smart grid network and is publicly available.

In this paper we address the issue of network cost efficiency for live streaming peer-to-peer systems. We formalize this as an optimization problem, which features a generic cost function. The latter is appropriate to capture not only ISP-specific link weights, but also non-linear, congestion-dependent costs. Our main contribution is the introduction of the Implicit-Primal-Dual scheme for flow control in live streaming peer-to-peer systems. It is fully distributed in that it relies only on local state variable exchanges. Moreover, we show that at a fluid scale, combined with random linear network coding, it admits the cost optimal operating point as a fixed point. We also prove asymptotic boundedness of fluid trajectories for particular cost functions. We finally show via experiments that these optimality properties are resilient to operational constraints such as finite generation size and finite field size.

User profiling is a useful primitive for constructing personalized services, such as content recommendation. In the present work we investigate the feasibility of user profiling in a distributed setting, with no central authority and only local information exchanges between users. Our main contributions are: (i)~We propose a spectral clustering technique, and prove its ability to recover unknown user profiles with only few measures of affinity between users. (ii)~We develop distributed algorithms which achieve an embedding of users into a low-dimensional space, based on spectral transformation. These involve simple message passing among users, and provably converge to the desired embedding.

We consider the problem of determining the "closest", or best Internet host to connect to, from a list of candidate servers. Most existing approaches rely on the use of metric, or more specifically Euclidean coordinates to infer network proximity. This is problematic, given that network distances such as latency are known to violate the triangle inequality. This leads us to consider non-metric coordinate systems. We perform an empirical comparison between the "min-plus" non-metric coordinates and two metric coordinates, namely L-infinity and Euclidean. We observe that, when sufficiently many dimensions are used, min-plus outperforms metric coordinates for predicting Internet latencies. We also consider the prediction of "widest path capacity" between nodes. In this framework, we propose a generalization of min-plus coordinates. These results apply when node coordinates consist in measured network proximity to a random subset of landmark nodes. We perform empirical validation of these results on widest path bandwidth between PlanetLab nodes. We conclude that appropriate non-metric coordinates such as generalized min-plus systems are better suited than metric systems for representing the underlying structure of Internet distances, measured either via latencies or bandwidth.