CTN Issue: July 2013

1. LACAS: Learning Automata-Based Congestion Avoidance Scheme for Healthcare Wireless Sensor Networks

One of the major challenges in deployed wireless sensor networks (WSN) is to curb down congestion in network’s traffic without compromising the energy consumption of the sensor nodes. Congestion disrupts the continuous flow of data, increases loss of information, delays data delivery to the destination and significantly and unnecessarily increases energy consumption in already energy-strapped nodes. Obviously, in healthcare WSN applications, particularly those that cater to medical emergencies or monitor patients in critical conditions, it is desirable to prevent congestion from occurring in the first place.

In this work, the authors address the problem of congestion in the nodes of healthcare WSN using a learning automata (LA)-based approach. The primary objective is to adaptively equate the processing rate (data packet arrival rate) in the nodes to the transmitting rate (packet service rate), so that the occurrence of congestion in the nodes can be avoided. The authors maintain that the proposed algorithm, named as Learning Automata-Based Congestion Avoidance Algorithm in Sensor Networks (LACAS), can counter the congestion problem in healthcare WSN effectively. An important feature of LACAS is that it intelligently “learns” from the past and improves its performance significantly as time progresses. The proposed LA-based model is evaluated using simulations representing healthcare WSN. The results obtained through the experiments with respect to performance criteria have important implications in the healthcare domain. For example, the number of collisions, the energy consumption at the nodes, the network throughput, the number of unicast packets delivered, the number of packets delivered to each node, the signals received and forwarded to the Medium Access Control (MAC) layer, and the change in energy consumption with variation in transmission range, have shown that the proposed algorithm is capable of successfully avoiding congestion in typical healthcare WSNs requiring a reliable congestion control mechanism.

Title and author(s) of the original paper in IEEE Xplore:
Title: LACAS: Learning Automata-Based Congestion Avoidance Scheme for Healthcare Wireless Sensor Networks
Author: S. Misra, V. Tiwari and M. S. Obaidat
This paper appears in: IEEE Journal on Selected Areas in Communications
Issue Date: May 2009

2. Scheduling Strategies to Mitigate the Impact of Bursty Traffic in Wireless Networks

Modern data networks support highly heterogeneous traffic sources. It is therefore imperative to design network control policies that are inherently robust to burstiness in traffic. Ideally, the control policies should not adversely affect the stability and delay properties of one traffic flow, due to erratic or bursty behavior of another traffic flow in the network.

Maximum weight scheduling (MWS) is a well-known link scheduling algorithm that achieves the maximum throughput in communication networks. Loosely speaking, the algorithm schedules the set of feasible links with the maximum total queue length at that instant. However, when there is a mix of bursty and benign traffic, MWS tends to allocate most of the service to the bursty traffic flow, following the arrival of a large burst, which in turn leads to large delays for the benign flows. Thus, bursty flows in a network can ‘infect’ benign flows with large erratic delays.

The main contribution of this paper is in designing scheduling mechanisms that achieve maximum throughput, and yet provide a degree of robustness against the effect of bursty traffic in the network. Specifically, we discuss two scheduling algorithms. The first algorithm is based on adaptive CSMA (Carrier Sense Multiple Access), a recently-developed randomized algorithm that achieves the maximum throughput in a distributed fashion. The second algorithm is maximum weight scheduling with capped queue lengths. Unlike the traditional MWS, this algorithm schedules the set of links with the maximum sum of capped queue lengths, where the cap value is chosen to be sufficiently large.

To characterize the performance of the algorithms, we consider a simple queuing network consisting of two conflicting links. The traffic served by the first link is bursty, modeled by a heavy-tailed arrival process, while traffic at the second link is benign, modeled by a light-tailed arrival process. In this setting, previous work has shown that even the light-tailed traffic would experience heavy-tailed delays under MWS. In contrast, we demonstrate a threshold phenomenon in the relationship between the arrival rate of the light-tailed traffic and its queue backlog distribution. In particular, we show that with the adaptive CSMA algorithm, the light-tailed traffic experiences a light-tailed queue backlog, when its arrival rate is less than a certain threshold value. Intuitively, this implies that the benign traffic generally experiences predictable delays. For arrival rates above the threshold, the light-tailed traffic experiences a heavy-tailed queue backlog, implying that the delays experienced can be large and highly erratic. Our analysis also shows that this threshold value is approximately equal to half the server capacity.

Intuitively, all links in adaptive CSMA attempt to capture the channel with bounded aggressiveness. As a result, even when large bursts arrive at a link carrying heavy-tailed traffic, the link cannot take over the server by attempting to transmit with arbitrary aggressiveness. This has the effect of 'shielding' the light-tailed traffic from the large bursts, at least when the arrival rate is smaller than the threshold value. Furthermore, adaptive CSMA does not need any a priori information about traffic statistics.

Note that for general queuing networks with more than two links and any mix of heavy-tailed and light-tailed traffic, the 'shielding' effect still holds due to the bounded aggressiveness, although the threshold arrival rates may be different from the two-link network.

Motivated by the above properties of adaptive CSMA, we show a similar threshold behavior for max-weight scheduling with capped queue lengths.

This paper wins a Best Paper Award in WiOPT 2013 conference.

Title and author(s) of the original paper in IEEE Xplore:
Title: Scheduling Strategies to Mitigate the Impact of Bursty Traffic in Wireless Networks
Author: K. P. Jagannathan, L. Jiang, P.L. Naik, and E. Modiano
This paper appears in: 11th International Symposium on Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks (WiOPT)
Issue Date: May 2013

Statements and opinions given in a work published by the IEEE or the IEEE Communications Society are the expressions of the author(s). Responsibility for the content of published articles rests upon the authors(s), not IEEE nor the IEEE Communications Society.