Call Admission Control in Wireless Ad Hoc Networks

Mobile ad hoc networks are used in wireless local area networks (WLAN) in the IEEE 802.11 standard. An ad hoc network is composed of a set of autonomous nodes created in a ubiquitous setup  with each node being an autonomous node. Any node can register with the network and use the medium to transmit packets to its peer node and can deregister at an arbitrary time. The distributed and autonomous nature of ad hoc networks might result in a futile performance due to over-utilization of the network resources by uncoordinated nodes. The lack of a central controller will accentuate the problem and if proper actions are not taken, the behavior of network might be driven below the acceptable threshold.

In a wireless ad hoc network, all backlogged nodes will compete with other active nodes to send their packets. In a non-hostile environment, all nodes should cooperate in bandwidth allocation and resource partitioning. However, the distributed nature of management process in an ad hoc network may allow destructive operation of malicious sources. Sources can be very greedy and might consume all available resources. A distributed call admission control will alleviate such an instance of network operation. In this research, we propose a measurement-based call admission controller for ad hoc networks.  The measurement-based approaches are divided into active and passive techniques and have recently been proposed for call admission control in the Internet.  In a passive approach, the call admission controller will gather necessary information by simply listening to the medium. Based on this observation, the controller might accept or reject the requested call. In an active approach, the node estimates the status of network by transmitting a number of probing packets over an appropriate time interval.

We have proposed a distributed call admission controller using the service-curve network provisioning approach. The service curve of a connection is defined as the amount of service given to that connection over a backlogged interval. We use a sequence of small-size probing packets to estimate the service curve of the network.  The estimated service curve is then used to devise a call admission controller. The admission of a new call should  maintain the performance of the network in an acceptable region. The boundary of this region is represented by the  universal service curve. The universal service curve is a linear curve independent of the number of sources and their activity and simply reflects the worst-case behavior of network. In the proposed technique, a call is accepted into the network if the admission of the new call will not drive the service curve below the universal service curve. If all sources abide by this rule, the overall performance will be lower bounded by the universal service curve. The universal service curve is then a non-random common reference curve that can be used to synchronize the whole network --- in terms of the worst-case QoS parameters --- and can be used by all nodes for call admission control and performance evaluation.

In the proposed call admission control, the input traffic is decomposed into conforming and nonconforming packets. The transmission of a conforming packet keeps the service curve above the universal service curve and the transmission of a nonconforming packet moves the service curve below the universal service curve. We will accept the call if the ratio of nonconforming traffic to the whole traffic is small.  The service curve approach can also be used for traffic shaping. We have shown that if an appropriate amount of delay is added to a nonconforming packet, it will become conforming. This suggests that the projected flow can be regulated at the transmitter to create conforming traffic. In fact, using the service curve provisioning approach will create  a mechanism that can smoothen bursty traffic and improve the network performance. 
 

Further Reading:

[1] S. Valaee and B. Li, “Distributed Call Admission Control for Ad Hoc Networks”, in IEEE Vehicular Technology Conference (VTC), Vancouver, BC, Canada, September 2002. [pdf]