Yu Wang
Email: ywang@cse.ucsc.edu
URL:
http://www.cse.ucsc.edu/~ywang
An ad hoc network is a dynamic network formed on demand by a group of nodes without the aid of any pre-existing network infrastructure. An efficient and effective medium access control (MAC) protocol which regulates nodes' access to the shared channel(s) is essential in an ad hoc network. Our work is focused on throughput and fairness properties of the existing omni-directional MAC protocols as well as enhancement of their performance with directional antennas via both analytical and simulation approaches.
In [1-3], we present the first analytical model to derive the saturation throughput of four-way sender-initiated collision avoidance protocols in multi-hop ad hoc networks. We show that the sender-initiated collision-avoidance scheme achieves much higher throughput than the idealized carrier sense multiple access scheme with an ideal separate channel for acknowledgments. More importantly, we show that the collision avoidance scheme can accommodate much fewer competing nodes within a region in a network infested with hidden terminals than in a fully-connected network, if reasonable throughput is to be maintained. Simulation experiments of the popular IEEE 802.11 MAC protocol validate the predictions made in the analysis.
In [4], we apply the aforementioned model to analyze three collision-avoidance protocols that use either all omni-directional transmissions, directional transmissions only, or a combination thereof. The analysis shows that collision avoidance using a narrow antenna beamwidth for the transmission of all control and data packets achieves the highest throughput among the three collision avoidance schemes considered. Simulation experiments validate the results predicted in the analysis. The results also show that transmissions with narrow beamwidth can also reduce the average delay experienced by nodes. It is concluded that the advantage of spatial reuse achieved by narrow beamwidth transmissions outweighs that of conservative collision avoidance schemes.
In [5], the tradeoff between spatial reuse and collision avoidance is investigated in detail and it is shown that in directional antennas enabled ad-hoc networks, the aggressive channel access scheme featured by all directional transmissions indeed outperforms other conservative schemes in terms of enhanced throughput and reduced delay.
In [6], we study the fairness with which competing flows share the channel in ad hoc networks using collision avoidance protocols. It is shown that the required multi-hop coordination makes the backoff-based distributed fair queueing schemes less effective. Using extensive simulations of two competing flows with different underlying network configurations, it is shown that the commonly used flow contention graph is insufficient to model the contention among nodes and that various degrees of unfairness can take place. The fairness problem is more severe in TCP-based flows due to the required acknowledgment traffic, and TCP throughput is also negatively affected. A measurement-based fair scheme is analyzed in which nodes estimate their fair share of the channel from overheard traffic and adjust their backoff window accordingly (voluntarily); it is shown that such a scheme achieves much better fairness but sacrifices too much throughput.
In [7-9], we propose a novel hybrid channel access scheme that combines both sender-initiated and received-initiated collision-avoidance handshake. The new scheme is compatible with the popular IEEE 802.11 MAC protocol and involves only some additional queue management and book-keeping work. Simulation experiments show that the new scheme can alleviate the fairness problems existent for both UDP and TCP based applications with almost no degradation in throughput. However, similar to the measurement-based fair scheme, it does not exchange information among nodes explicitly and hence cannot solve the fairness problem conclusively. Based on the work done in [6-9], we conclude that more explicit information exchange among contending nodes is mandatory to solve the fairness problem conclusively while maintaining reasonable throughput.
In our ongoing and future work, we will focus on the following topics. At first, we will explore the use of directional antennas as another dimension in the solution space that can help to improve the performance of broadcast in ad hoc networks. Secondly, we will explore more realistic analytical models for MAC protocols using directional antennas. In the first step, we will consider the difference between omni-directional and directional transmission ranges and powerful antenna system that is also capable of directional receiving. In the second step, we will consider the influence of side lobes in directional transmissions and receptions on performance. Thirdly, we will investigate how the use of directional antennas can help to improve the fairness %and energy efficiency of MAC protocols for ad hoc networks and come up with schemes that have better fairness while maintaining reasonable throughput.