WMPG Technical Proposal
Internet Architecture White Papers
(Phy-Link) Layers in the Brave New World
Recent years have seen an explosion in the use of wireless devices at the edge of the network. The protocol stacks have largely treated the wireless links as "equivalent" to,
possibly error-prone, wired links. While such an approach has served the purpose so far, it appears that the this approach may not suffice in the future. Two trends are of particular interest:
- Rise of "cooperative" wireless networks: While in a wired network, typically, a "link" is fully defined by its two end-points, in wireless networks, the notion of a "link" may have to be revised. Some examples:
Such approaches allow the use of multiple "hops" to deliver packets over what may be construed to be a single "link".
Present link layer abstractions may not be sufficient for these more general notions of a link.
- Ad hoc networks: In such networks, hosts route packets on each other's behalf, requiring the use of multi-hop routing among the hosts to deliver packets. Potentially, the "routing" functionality can be implemented at the link layer, such that the entire ad hoc network appears as a single "link" from the perspective of the rest of the network.
- Cooperative diversity: With cooperative diversity, hosts other than the end-points of a "link" cooperate to help deliver packets on the link. For instance, when A transmits a packet to B, a third host C may relay a function of the "packet" received from A to C; host C then can combine the receptions from A and C both to determine the received packet.
- Network coding: Network coding allows hosts to relay not just a received packet, but instead a function of packets received from different sources.
- Adaptation mechanisms: Due to the limited capacity of wireless networks, a large range of adaptation mechanisms - facilitated by various forms of diversity - have been explored. Some examples include power and rate control, channel adaptation, and directional beamforming. While such adaptations have the potential for delivering greater performance, they also muddle the notion of a
"link". For instance, given that a host can transmit at two different power levels, should a "broadcast" be performed at the lower power level, or the higher power level? Should the broadcast be performed on any one available channel, or all available channels? In general, the answer to such questions will have to depend on the higher layer's requirements, making the adaptations necessarily cross-layer. Suitable interfaces need to be developed to allow for such cross-layer interactions. Similarly, given that the physical layer behavior is defined by many parameters (e.g., rate, power, channel, beamform, etc.), the network layer may want to specify the parameters (or parameter range) to be used for transmitting each packet, based on network level information (as opposed to local information available to each host).
One simple approach to avoid the above complexities would be to not use such optimizations or cooperative mechanisms. This would result in performance degradation, however. When the available capacity is limited, such degradation is not likely to be acceptable. However, if other approach dramatically increase the capacity of wireless networks (e.g., using larger amounts of spectrum), then it may become desirable to trade-off some performance to obtain simplicities in protocol design. In the meantime, however, we suggest that approaches be developed to facilitate cross-layer cooperation, particularly at the lower layers of the protocol stack, in a manner that is "future-proof".
Instructions for Workshop Participants
Directions to WINLAB Tech Center II facility
Directions to Workshop Dinner on Aug 2