Event-Based System Architecture in Mobile Ad Hoc Networks (MANETs)

Introduction

The publish/subscribe paradigm (Muhl, Fiege, & Pietzuch, 2006) communicates on the basis of either the message content or the message source being of interest to destinations – as opposed to the source specifying the recipient(s). P/S systems can be considered to be a form of event-based systems, in the sense that the information injected to and propagated through the system can be treated as events. A unit in the system can act either/both as information producers (publishers) or/and consumers (subscribers). Subscribers declare their interests via subscriptions to certain events, most commonly specified by the content or the topic of the events (with different expressive power), and publishers produce events of information to the system. The event routing mechanism implemented in the P/S system (usually middleware) then takes charge of the event delivery according to the subscription knowledge.

The strong decoupling between information producers’ and consumers’ identities in a P/S system is appealing in loosely coupled network background such as mobile ad hoc networks (MANETs) (Corson & Macker, 1999), because of the ease with which components can be added, removed or changed at runtime. What is more, P/S features the capability of supporting one-to-many connectivity along with redundant information producers and consumers. This is conducive to constructing fault-tolerant or high-availability applications with redundancy and fault detection/handling mechanisms, and also building reconfigurable applications in a dynamic or even intermittent environment (Castellote & Bolton, 2002).

A MANET is a collection of mobile devices with dynamically changing membership and multi-hop topologies composed of wireless links. By its nature, a MANET is a self-organizing adaptive network, and thus needs to be formed and maintained in a distributed manner without centralized support or fixed infrastructures (Sarkar, Basavaraju, & Puttamadappa, 2007).

However, the potential advantages of the P/S interaction model atop MANETs are not fully realized by the state of the art predominant products and research solutions, such as TIB/RV (Oki, Pfluegel, Siegel, & Skeen, 1993), SCRIBE (Castro, Druschel, Kermarrec, & Rowston, 2002), SIENA (Carzaniga, Rosenblum, & Wolf, 2001), REDS (Cugola, Nitto, & Fuggetta, 1998), Kyra (Cao & Singh, 2004), IBM’s WS-Notification (Graham, Niblett, Chappell, Lewis, Nagaratnam, Parikh, Patil, Samdarshi, Sedukhin, Snelling, Tuecke, Vambenepe, & Weihl, 2004), RTI’s DDS (Castellote & Bolton, 2002) based on OMG's DDS (Object Management Group, 2007).

To fully and yet carefully explore the applicability of publish/subscribe systems overlaid on MANETs, one must cope with network uncertainties and constraints, including (but not limited to):

• •

  frequent link breakages and temporary network disconnections;

• •

  temporary node unavailability and node joins or departures at unpredictable times; and

• •

  mobility-induced resource constraints on the overall architecture, such as limits on bandwidth, latency, and energy consumption.

Some efforts have been made (Fiege, Gartner, Kasten, & Zeidler, 2003; Cao & Shen, 2007; Muthusamy, Petrovic, & Jacobsen, 2005) assuming either that only a subset of nodes in the network can roam and act as clients or that the clients are always at one-hop away from the fixed infrastructure, and therefore focus only on a restricted subset of the problem space. A more general and common scenario in MANET is that every node in the network has mobility and can access the publish/subscribe service (e.g., by running a publish/subscribe middleware), while also acting as brokers for message forwarding and event matching to make the service available.