Search

Throughput optimal scheduling policies in general require the solution of a complex and often NP-hard optimization problem. Related literature has shown that in the context of time-varying channels, randomized scheduling policies can be employed to reduce the complexity of the optimization problem but at the expense of a memory requirement that is exponential in the number of data flows. In this paper, we consider a Linear-Memory Randomized Scheduling Policy (LM-RSP) that is based on a pick-and-compare principle in a time-varying network with $N$ one-hop data flows. For general ergodic channel processes, we study the performance of LM-RSP in terms of its stability region and average delay. Specifically, we show that LM-RSP can stabilize a fraction of the capacity region. Our analysis characterizes this fraction as well as the average delay as a function of channel variations and the efficiency of LM-RSP in choosing an appropriate schedule vector. Applying these results to a class of Markovian channels, we provide explicit results on the stability region and delay performance of LM-RSP., Long version of preprint to appear in the IEEE Transactions on Networking

Recently wireless sensor networks featuring direct sink access have been studied as an efficient architecture to gather and process data for numerous applications. We focus on the joint effect of clustering and data correlation on the performance of such networks. We propose a novel cluster-based data collection scheme for sensor networks with direct sink access (CDC-DSA), and provide an analytical framework to evaluate its performance in terms of energy consumption, latency, and robustness. In our scheme, CHs use a low-overhead and simple medium access control (MAC) conceptually similar to ALOHA to contend for the reachback channel to the data sink. Since in our model data is collected periodically, the packet arrival is not modeled by a continuous random process and, therefore, our framework is based on transient analysis rather than a steady state analysis. Using random geometry tools, we study how the optimal average cluster size and energy savings vary in a response to various data correlation levels under the proposed MAC. Extensive simulations for various protocol parameters show that our analysis is fairly accurate for a wide range of parameters. Our results suggest that despite the tradeoff between energy consumption and latency, both of which can be substantially reduced by proper clustering design.

In this paper, we consider CSMA policies for scheduling of multihop wireless networks with one-hop traffic. The main contribution of this paper is to propose Unlocking CSMA (U-CSMA) policy that enables to obtain high throughput with low (average) packet delay for large wireless networks. In particular, the delay under U-CSMA policy becomes order-optimal. For one-hop traffic, delay is defined to be order-optimal if it is O(1), i.e., it stays bounded, as the network-size increases to infinity. Using mean field theory techniques, we analytically show that for torus (grid-like) interference topologies with one-hop traffic, to achieve a network load of $\rho$, the delay under U-CSMA policy becomes $O(1/(1-\rho)^{3})$ as the network-size increases, and hence, delay becomes order optimal. We conduct simulations for general random geometric interference topologies under U-CSMA policy combined with congestion control to maximize a network-wide utility. These simulations confirm that order optimality holds, and that we can use U-CSMA policy jointly with congestion control to operate close to the optimal utility with a low packet delay in arbitrarily large random geometric topologies. To the best of our knowledge, it is for the first time that a simple distributed scheduling policy is proposed that in addition to throughput/utility-optimality exhibits delay order-optimality., Comment: 44 pages

We consider CSMA policies for multihop wireless networks. CSMA policies are simple policies that can be easily implemented in a distributed manner. However, the delay performance of CSMA policies can be very poor as the delay can grow exponentially in the network size. As a result, CSMA policies are not practical for delay-sensitive traffic even for mid-sized networks. In this paper, we consider a slight variant of the classical CSMA policies and show that it leads to a much improved delay performance. In particular, we show that the delay does not depend on the networks size. Using this result, we also characterize the delay-throughput trade-off of the proposed CSMA policy. At the heart of our analysis is a result that shows that CSMA policies quickly converge to a maximum schedule, i.e., converge to a maximum schedule at a rate that does not depend on the network size. Using this insight, we consider a CSMA policy that periodically “unlocks” the transmission pattern of a CSMA policy, and show that this unlocking mechanism can be used to obtain a much improved delay performance without significantly reducing the throughput. While our analysis has been carried out for the special case of an interference graph with a grid (lattice) topology, we provide numerical case studies for general network topologies and show that the intuition obtained from the analysis carries over to these general cases. We also illustrate the performance of the proposed CSMA policy when combined with a flow control mechanism.

It is well known that for ergodic channel processes the Generalized Max-Weight Matching (GMWM) scheduling policy stabilizes the network for any supportable arrival rate vector within the network capacity region. This policy, however, often requires the solution of an NP-hard optimization problem. This has motivated many researchers to develop sub-optimal algorithms that approximate the GMWM policy in selecting schedule vectors. One implicit assumption commonly shared in this context is that during the algorithm runtime, the channel states remain effectively unchanged. This assumption may not hold as the time needed to select near-optimal schedule vectors usually increases quickly with the network size. In this paper, we incorporate channel variations and the time-efficiency of sub-optimal algorithms into the scheduler design, to dynamically tune the algorithm runtime considering the tradeoff between algorithm efficiency and its robustness to changing channel states. Specifically, we propose a Dynamic Control Policy (DCP) that operates on top of a given sub-optimal algorithm, and dynamically but in a large time-scale adjusts the time given to the algorithm according to queue backlog and channel correlations. This policy does not require knowledge of the structure of the given sub-optimal algorithm, and with low overhead can be implemented in a distributed manner. Using a novel Lyapunov analysis, we characterize the throughput stability region induced by DCP and show that our characterization can be tight. We also show that the throughput stability region of DCP is at least as large as that of any other static policy. Finally, we provide two case studies to gain further intuition into the performance of DCP., Submitted for journal consideration. A shorter version was presented in IEEE IWQoS 2008

Throughput optimal scheduling policies in general require the solution of a complex optimization problem. The past literature has shown that the complexity of this optimization problem can be greatly reduced, but at the expense of memory requirement that is exponential with the number of users. In this paper, we study the stability region of a class of linear-memory scheduling policies for time varying channels, and investigate how the channel memory impacts the supportable input rates. The set of scheduling policies in this paper covers a wide spectrum of resource allocation algorithms, which allows us to study policies with different complexity levels. In particular, we are able to model a class of low-complexity scheduling policies with linear memory, which are suitable for practical implementation.

In wireless sensor networks, clustering allows the aggregation of sensor data. It is well known that leveraging the correlation between different samples of the observed data will lead to better utilization of energy reserve. However, no previous work has analyzed the effect of non-ideal data aggregation in multi-hop sensor networks. In this paper, we propose a novel analytical framework to study how partially correlated data affect the performance of clustering algorithms. We analyze the behavior of multi-hop routing and, by combining random geometry techniques and rate distortion theory, predict the total energy consumption and network lifetime. We show that when a moderate amount of correlation is available, the optimal probabilities that lead to minimum energy consumption are far from optimality in terms of network lifetime. In addition, we study the sensitivity of the total energy consumption and network lifetime to the amount of correlation and compression distortion constraint.