Publications (2)0 Total impact
Conference Paper: An interaction between network coding and end-host coding[Show abstract] [Hide abstract]
ABSTRACT: Network coding techniques have been proved effective in increasing the capacity of wireless ad hoc and mesh networks. Despite this, little is known about its potentials and limitations of improving application perceivable performance. One of the mysteries lies in the complicated interaction between the low-layer network coding function and the upper-layer protocols/applications. In this paper, targeting multimedia applications, we attempt to inspect the interaction between network coding and end-host coding techniques, in particular forward error correction (FEC), and study how network coding can benefit multimedia applications by interacting with FEC. We show network coding has two positive impacts on the efficacy of FEC. First, when network capacity is critically low for competing multimedia flows, even a marginal capacity increase leads to much higher packet recovery ratio for applications. Second, we show that network coding technique can lead to less bursty packet loss patterns. Even at a fair loss rate, such less bursty loss patterns lead to dramatically higher packet recovery ratio at the receivers. Our results and analysis validate that network coding can be beneficial to end-host coding techniques.
Conference Paper: Diagnosing the limitations of network coding at transport layer[Show abstract] [Hide abstract]
ABSTRACT: The recent development on network coding techniques has attracted much attention, with their demonstrated potentials in improving network capacity and reliability, particularly in wireless networks with broadcast capability. While most studies have focused on designing efficient algorithms and protocols to exploit the benefits at lower layers, i.e., the link layer and network layer, we reveal the impacts of network coding on transport layer algorithms. We show that transport layer protocols, such as TCP and UDP, can suffer from the less careful design of network coding algorithms at lower layers. For example, the behavior of TCP flows and its congestion control algorithms is less predictable, when the underlying network is enriched with the network coding capability. We argue that this added capability changes the fundamental assumptions on the networks and the demographics of the packet flows, and thus adversely impacts the end-to-end performance of the applications. We propose approaches to alleviate the impacts of network coding on upper-layer protocols.
Case Western Reserve UniversityCleveland, Ohio, United States