Realizing the benefits of user-level channel diversity

Abstract

Channel or path diversity is known to improve performance in physical layer designs, channel access strategies, path switching mechanisms, etc. In this paper, we focus on "user-level" mechanisms that operate simply by distributing packet transmissions across multiple channels. We seek to understand when, why, and to what extent this can be of benefit, and equally important, whether these benefits can be realized with as little of an added cost as possible. In that context, our main contribution is not so much in identifying optimal policies for leveraging channel diversity, but in introducing the concept of channel "equivalence" and demonstrating that channel diversity yields substantial benefits mostly when channels are approximately equivalent. We build on this finding to investigate the robustness of these improvements against errors in the characterization of the available channels or changes in their characteristics. We also explore the sensitivity of the results as the number of available channels varies. The findings of the paper demonstrate that by allowing packet transmissions from multiple users to intelligently share channels, it is possible to improve overall performance and robustness through simple and portable user-level mechanisms.

Exploiting the Path Propagation Time Differences in Multipath Transmission with FEC

“…For instance, in the inset in Fig. 7 we show the performance of Spread under various rate configurations (n 1 , n− n 1 ); the minimum is reached for (7,3). As descried in IV-C, allowing for this rate optimization leads to the optimal Spread schedule S Additionally, the dashed curve shows the effective loss rate of the optimal schedule, where packets are not restricted to even spacing on each path, as described in Section IV.…”

“…We denote by Immediate the schedule S imm = (T imm , R imm ) that represents the approach used in [4]- [7,9,10]. As the name suggests, Immediate sends the data packets as soon as they are generated, i.e., every time interval T .…”

“…. .=n R , then this boils down to a simple round-robin schedule applied in [4,6,7,9]. In contrast, when the rates differ, a more elaborate approach should be used.…”

“…optimal Immediate multipath schedule S imm opt uses mainly (or only) the better path, which substantially limits the gain of multipath [5,7]. This is illustrated in plot (D) by the dashed curve; the ratio π * B ≈ 0.25 the path P 2 becomes too lossy, and both Immediate and Spread send all packets on P 1 only and thus become equivalent.…”

“…It got more attention recently, e.g., in [4]- [7,9,10]. Multipath was also studied in the context of Multiple Description Coding [11].…”