Monday, February 26, 2007

Multichannel MAC Protocols for Wireless Networks.

Multichannel MAC Protocols for Wireless Networks Ritesh Maheshwari, Himanshu Gupta, Samir Das, In Proc of IEEE SECON 2006, Reston, VA, Sept 2006

This paper I don't think I saw at the conference, even though I attended and presented a paper. I had to leave early to go to another conference, maybe that's how I missed it.

Anyhow, now I caught up with it, and it is one of these papers that are nice to read in the morning as you don't need to much caffeine to grasp what it is about: the papers present two protocols to access multiple channel with a single transmission interface (and in one case, a busy tone interface). The protocols are intuitive enough, the motivations are clear, the validation show the promised improvement. There is no analysis, just a simulation validation, so it is 8 pages of light reading.

The two protocols that are described are some variation of other protocols, which is interesting to provide some context for these other protocols. The two protocols they improve upon are RDT and MMAC. MMAC is a protocol which uses a mechanism similar to the power save mechanism of 802.11 to define a control time slot and a transmission phase. The control slot is used for the nodes to agree on which channel to switch to for transmission. Nodes should be synchronized to know when the control time slot is and when the transmission phase ends. All nodes meet on the same channel during the control slot.

RDT (Receiver Directed Transmission) is a protocol for which each nodes select a "quiescent channel", which is the channel where it will receive transmission. It switches to the quiescent channel of the receiver when it wishes to transmit to that receiver. Allocation of quiescent channel can be done off-line or dynamically.

One more protocol is mentioned in the paper: DCA. DCA (Dynamic Channel Assignment) is a two-interface protocol: one interface is the control interface, used by the nodes to schedule to which channel they will switch their other interface, which they use for the actual transmission.

Ok: that's the background. The contribution: the paper introduces a version of xRDT which is extended with a busy tone interface (a simpler-to-implement interface) and an explicit notification of the termination of the transmission, so as to reduce the deafness of the terminals (a terminal is deaf when not on his quiescent channel; a would-be transmitter needs to explicitly know when his intended receiver returns to its quiescent channel).

The paper also introduces another protocol, Local Coordination-based Multichannel (LCM)MAC, which works on a single interface and is a self-clocking (ie. nodes synchronize themselves) version of MMAC. Nodes define a transmission schedule which include a control phase and a data transmission phase. The first one to win the race to set-up the schedule wins and the other act according to that schedule.

Performance analysis is straightforward. The only odd part is that RDT is not included (DCA, MMCA and the two new protocols, xRDT and LCM-MAC are). I would assume xRDT offering an improvement on RDT would be subject to validation. Other than that, it is only a simulation, which shows the benefit of multiple interfaces over single interface, and the benefit of the new protocols over the legacy protocols.

No comments: