A Result on Hybrid Scheduling in Wireless Networks

Medium access control is an exceedingly important aspect of the operation of any network. This is particularly true in wireless settings where the broadcast nature of the medium accentuates the need for effective medium access algorithms which provide good throughput and fairness characteristics. By far the most popular MAC techniques for wireless networks are those based on Carrier Sense Multiple Access (CSMA) due to their simplicity and amenability to distributed implementation. The IEEE 802.11 MAC is an example of a widely used MAC protocol that uses CSMA. However, CSMA protocols are often inadequate when fairness is of importance. Moreover, the performance of CSMA protocols degrades when the number of active transmitters increases since the probability of collisions increases. In such scenarios, Time Division Multiple Access (TDMA) based approaches are much more suitable. In TDMA protocols, each user receives a pre-allocated time-slice during which it can access the channel. This provides for fair distribution of bandwidth across users. However, when only a few users are active, TDMA leads to waste of bandwidth, as a time-slice will go unutilized if the owner of that time-slice has no tarffic to send. Evidently, it is desirable for a MAC procedure to have the good characteristics of both classes of MAC techniques, i.e., it should provide fair medium access when there are many active users, but should avoid wastage of bandwidth when only a few users are active. This has lead to the design of hybrid TDMA-CSMA approaches, e.g. Z-MAC. However, there has been little effort toward developing formal theoretical results to examine and understand such hybrid strategies. At the same time, there has been much recent work in the theoretical domain on designing distributed scheduling algorithms that can schedule traffic while maintaining finite (expected) queues. An important class of scheduling algorithms that has received much attention lately are the so-called maximal schedulers. The performance of a maximal scheduler with a local threshold rule is studied in [1], [2], while the rate-stability of certain prioritized maximal schedulers is considered in [3]. Approximations to maximal schedulers can potentially be implemented using backoff mechanisms in conjunction with CSMA [4], [5]. It is also to be noted that some characteristics of CSMA based medium access protocols, e.g., 802.11, are approximately similar to those of a maximal scheduler, since a node that has data to send will try to access the channel and unless it detects the channel to be occupied by another interfering transmission it will gain access to it after executing the specified medium access procedure.1 Thus, maximal scheduling provides a tractable theoretical abstraction that can assist in building intuition. Here, we present a performance bound for a hybrid TDMA-Maximal scheduler.