Survivability and performance optimization in communication networks using network coding

Publication Year:
2015
Usage 145
Downloads 110
Abstract Views 35
Repository URL:
https://lib.dr.iastate.edu/etd/14513
DOI:
10.31274/etd-180810-4062
Author(s):
Mohandespour, Mirzad
Publisher(s):
Iowa State University
Tags:
Computer Engineering; Communication networks; Network coding; Performance optimization; Survivability
thesis / dissertation description
The benefits of network coding are investigated in two types of communication networks: optical backbone networks and wireless networks. In backbone networks, network coding is used to improve survivability of the network against failures. In particular, network coding-based protection schemes are presented for unicast and multicast traffic models. In the unicast case, network coding was previously shown to offer near-instantaneous failure recovery at the bandwidth cost of shared backup path protection. Here, cost-effective polynomial-time heuristic algorithms are proposed for online provisioning and protection of unicast traffic. In the multicast case, network coding is used to extend the traditional live backup (1+1) unicast protection to multicast protection; hence called multicast 1+1 protection. It provides instantaneous recovery for single failures in any bi-connected network with the minimum bandwidth cost. Optimal formulation and efficient heuristic algorithms are proposed and experimentally evaluated. In wireless networks, performance benefits of network coding in multicast transmission are studied. Joint scheduling and performance optimization formulations are presented for rate, energy, and delay under routing and network coding assumptions. The scheduling component of the problem is simplified by timesharing over randomly-selected sets of non-interfering wireless links. Selecting only a linear number of such sets is shown to be rate and energy effective. While routing performs very close to network coding in terms of rate, the solution convergence time is around 1000-fold compared to network coding. It is shown that energy benefit of network coding increases as the multicast rate demand is increased. Investigation of energy-rate and delay-rate relationships shows both parameters increase non-linearly as the multicast rate is increased.