High performance publish/subscribe protocols for wireless ad hoc networks using geographic information

Abstract

Wireless Ad hoc networks are emerging network paradigm which can be self-organized without any aids from the pre-established infrastructures. The networks have been widely used to support distributed, wireless, and mobile computing and communication applications. One of the major tasks of wireless ad hoc networks is to provide event service to the applications over the networks. Using event services, the applications can focus on producing and consuming events other than the transmitting and receiving events. Publish/Subscribe (pub/sub) is one of the most widely used middleware paradigm for event service in traditional distributed systems. However, wireless ad hoc networks have many unique characteristics different from traditional wired networks in terms of communication, mobility and resource constraints. These characteristics make the design of the pub/sub to be much more challenging. In this thesis, we investigate how to design efficient publish/subscribe (pub/sub) protocols for wireless ad hoc networks. We review the existing pub/sub protocols designed for wireless ad hoc networks. Based on the review of previous studies, we proposed a 3-D framework for protocol design in wireless ad hoc networks. We describe three novel pub/sub protocols designed, respectively, for wireless sensor networks (WSNs), wireless mesh networks (WMNs), delay-tolerant networks (DTNs). The proposed protocols take the advantage of the geographical information of the network nodes to achieve high performance in event subscription and delivery. First, we describe HESPER, a Highly Efficient and Scalable Publish/subscribe protocol designed for wireless sEnsoR networks. HESPER uses the geographical information of the interesting events to efficiently disseminate subscriptions without flooding and establish the event delivery paths shared by subscribers without the coordination among the subscribers. Thus, HESPER can reduce the amount of traffic for event delivery and, thus significantly save energy of WSNs. To avoid rapid exhaustions of the nodes on the delivery paths, HESPER uses the rotation of intermediate nodes to balance the working loads for the nodes on the delivery paths. In addition, HESPER offers two different strategies to balance the tradeoff between the subscription and publication costs to satisfy the different requirements of the practical applications. Then, we propose a novel PUb/sub protocol for wireless Mesh networks based on locAtion prediction, namely PUMA, which can reliably and efficiently deliver events to mobile clients in a continuous manner. Using the location prediction for a mobile client, PUMA can deliver the interesting events to the location where a mobile client may stay. We show that the problem of delivery events to these locations with a minimum message cost, namely all-location event delivery problem, is a NP-complete problem. To further reduce the message cost, PUMA first disseminate the events to the locations, namely hot locations, where a mobile client has a higher probability to stay. We show that the problem of delivering an event to hot locations with a minimum message cost, namely hot-location event delivery problem, is also a NP-complete problem. We propose two algorithms for the both of two NP-complete problems. Thus, PUMA can guarantee the reliability of event delivery with low cost. Next, we propose a publish/subscribe (pub/sub) protocol, called GIANTs, that uses geographical information to improve the reliability and efficiency of event service for DTNs. Comparing with existing solutions, GIANT provides greater flexibility for applications in the sense that it allows recruiting randomly moving ferry nodes rather than only the ferries moving along specific directions. To guarantee the specified event delivery ratio, multiple ferries are needed to deliver the events, but this can lead to delivery redundancy. GIANT reduces the redundancy by first constructing an event delivery tree with optimal number of hops on the paths from the publishers to the subscribers, and then, using the location information of forwarding nodes to minimize the number of ferries to be recruited on each hop. Thus, in addition to flexibility and reliability, GIANT also achieves high efficiency.