Roger G. Kermode

Implementation experience with MANET routing protocols

This paper outlines our experience with the implementation and deployment of two MANET routing protocols on a five node, four hop, network. The work was prompted by the lack of published results concerning the issues associated with the implementation of MANET routing protocols on actual wireless networks, as opposed to results of simulation experiments. We examined implementations of two distance vector MANET routing protocols and found a number of problems with both protocols during the course of our experiments. The most significant was that neither protocol could provide a stable route over any multi-hop network connection. The route discovery process of both protocols is fooled by the transient availability of network links to nodes that were more than one hop away. Packets transmitted over a fading channel cause the routing protocol to conclude incorrectly that there is a new one hop neighbor that could provide a lower metric (hop count) route to even more distant nodes. This can occur even when nodes are stationary, mobility resulted in even less route stability. We implemented a simple signal strength based neighbor selection procedure to test our assertion that fading channels and unreliable network links were the cause of the failure of the routing protocols. The result was that neighbor discovery and the filtering for neighbors with which nodes could communicate reliably enables the creation of reliable multihop routes. Based on our experiences, we outline several recommendations for future work in MANET research.

Scoped hybrid automatic repeat reQuest with forward error correction (SHARQFEC)

Reliable multicast protocols scale only as well as their ability to localize traffic. This is true for repair requests, repairs, and the session traffic that enables receivers to suppress extraneous requests and repairs. We propose a new reliable multicast traffic localization technique called Scoped Hybrid Automatic Repeat reQuest with Forward Error Correction (SHARQFEC). SHARQFEC operates in an end-to-end fashion and localizes traffic using a hierarchy of administratively scoped regions. Session traffic is further reduced through the use of a novel method for indirectly determining the distances between session members. For large sessions, this mechanism reduces the amount of session traffic by several orders of magnitude over non-scoped protocols such as Scalable Reliable Multicast (SRM). Forward Error Correction is selectively added to regions which are experiencing greater loss, thereby reducing the volume of repair traffic and recovery times. Receivers request additional repairs as necessary. Simulations show that SHARQFEC out performs both SRM and non-scoped hybrid Automatic Repeat reQuest / Forward Error Correction protocols. Assuming the widespread deployment of administrative scoping, SHARQFEC could conceivably provide scalable reliable delivery to tens of millions of receivers without huge increases in network bandwidth.

A scalable multicast architecture for one-to-many telepresentations

We have developed a scalable reliable multicast architecture for delivering one-to-many telepresentations. In contrast to audio and video, which are often transmitted unreliably other media, such as slides, images and animations require reliability. Our approach transmits the data in two layers. One layer is for session-persistent data, with reliability achieved by FEC alone, using the Fcast protocol. The other layer is for dynamic data, with reliability achieved using the ECSRM protocol, which combines FEC with NACK suppression. Our approach is scalable to large heterogeneous receiver sets, and supports late-joining receivers. We have implemented our approach in a multicast version of PowerPoint, a graphical slide presentation tool.

Media banks: entertainment and the Internet

There are two emerging models for delivering high-density, synchronized audiovisual presentations: video-on-demand and the Internet. The first is based on long-lived connections and guaranteed timeliness; the second assumes short spurts of low-bandwidth data on demand. We present the design for a Media Bank that intermediates between them. It provides on-demand access to media elements that are assembled on-the-fly by the recipient to reproduce synchronized audiovisual presentations. The Media Bank uses a fully distributed architecture that assumes a community of viewers. Any member citizen can request or deliver sound, picture, descriptive annotations, or programs to control assembly. Data are redundantly stored in small segments and are cataloged by content and format to facilitate personalized and interactive retrieval.

Coding for content: enhanced resolution from coding

This paper describes a new video coding algorithm [Kermode 1994] that extends the traditional hybrid transform coding techniques used in MPEG in two ways. First, pictures are analyzed and segmented into a number of regions roughly corresponding to the background and any moving foreground objects. These regions are then encoded independently from one another thereby incorporating degree of useful content into the encoded representation beyond the pictures themselves. Second, the new algorithm also stores the background regions from several frames which can then be used to construct an artificial image at a higher resolution than that of the original background. Furthermore, both of these features are implemented at no overall cost in compression efficiency. In fact the new algorithm still outperforms MPEG, even for particularly degenerate sequences where the underlying assumption used for segmentation is broken.

Support for Reliable Sessions with a Large Number of Members

The ability to localize traffic when performing distributed searches within a group of nodes that form a session is a key factor in determining how big the group can scale. In this paper we describe an algorithm using the concept of scoping that we believe significantly enhances the ability to localize traffic for the service discovery aspect of many protocols, and hence their ability to scale. The algorithm is based upon the notion of a hierarchy of administrative multicast scopes where smaller scopes nest inside larger ones. To exploit this topological structure, we describe an application-layer protocol, the Scoped Address Discovery Protocol (SADP), which provides session members with the ability to discover, for each session, which addresses to use within each of the various scopes within a hierarchy. We show via simulation that SADP affords this ability in a manner that scales, through merging the well known distribution mechanisms of announce/listen and query/response and exploiting the nested hierarchy of scopes itself.

A novel method for video-on-demand via digital broadcast

We propose a new method for the on-demand delivery of streaming media over a small number of multicast channels to large numbers of receivers called Truncated Exponential Partitioning (TEP). Objects delivered using TEP are consumed from the beginning to the end in a linear fashion. This allows their data to be segmented and carried in different channels that are scheduled so that each can arrive just in time for presentation. As a consequence, the amount of storage required in the receiver, access latency, and bandwidth required for a desired level of performance are minimized. Quantitative expressions for these metrics are derived for TEP and compared to a similar scheme skyscraper broadcasting (SB) scheme from which the design of TEP was based. Simulations show that TEP offers superior performance to the SE scheme. Finally, an architecture for delivering movies over cable networks using TEP is proposed.