Mobile Ad-hoc Networks (MANETs) are characterized by constantly changing network topology and dynamic group membership. This makes implementing IP layer multicast in MANETs a challenging task. In recent years, several researchers have proposed the concept of overlay multicast (also known as application layer multicast). Even though application layer multicast is not as efficient as traditional IP based multicast, its flexibility in adapting to different environments and ease of implementation has contributed to its steady growth. Overlay multicast networks can be used to build roll-based priority trees. Such priority trees cannot be implemented at IP layer since IP multicast does not allow assigning priorities to different multicast sessions. It is easy to assign priorities to different multicast groups when multicast functionality is performed at the application layer.; In an ad-hoc wireless network, the position of nodes constantly changes; as a result, overlay multicast trees that are built using location information of member nodes will have a low latency. However, the performance gains of such trees are offset by the cost involved in maintaining precise location information. Periodic location updates generate a lot of overhead. As the degree of (location) accuracy increases, the performance improves but the overhead required to update and broadcast this information also increases. This dissertation proposes a design to build a sub-optimal location aided overlay multicast trees, where location updates of each member node are event based and the location broadcasts are limited to a small set of nodes. Since most ad hoc networks are implemented in indoor environment, this dissertation also looks at various indoor location sensing techniques. It closely examines three such indoor location sensing prototype systems - LANDMARC, Location sensing using Bluetooth and Bluebot. The LANDMARC and Bluetooth Location systems were implemented in the Experimental Laboratory for Advanced Networks and Systems (ELANS) lab at CSE-MSU, while the Bluebot system was designed and implemented at IBM T. J. Watson Research Center (Hawthorne, New York) as part of an internship project.; Finally, the dissertation examines the issue of resource allocation in multicast networks (esp. in MANETs). In a typical multicast network, a single tree is built with the source node as the root. In such a tree, only a few internal nodes contribute most of the resources and are involved in performing the multicast functionality. This leads to an uneven utilization of network resources. This problem is more prevalent in MANETs where network resources are limited. A possible solution to the problem is to split the multicast content over a number of trees. Multiple trees provide several paths for the multicast content and get more nodes involved in implementing the multicast functionality. However, in this setup, not all the trees get to use the best weight edges, thus the overall multicast latency increases. The dissertation examines MEST (Multiple Edge Sharing Trees), a distributed algorithm to construct multiple edge-sharing trees for small group multicast. MEST balances the resource allocation and delay constraints by choosing to overlap certain edges that have low weights.
展开▼
