Intelligent application persistence in tactical wireless networks

Abstract

Mobile tactical MANETs are subject to periods of intermittent connectivity and transient events of significant packet loss. Because of such disruptions, protocols and technologies have been proposed that are disruption and delay tolerant. One class of such systems works by storing messages/packets in the network. If connectivity to neighbors is lost, missing messages may be delivered when connectivity to neighbors is restored. This delay tolerant networking (DTN) technology aims to reliably deliver all the data that a source directs towards a receiver node. We suggest that in some cases it is of little utility and may even be counter-productive to persist and deliver all packets lost during a disruption episode. For example, in situational awareness applications, retransmitting many seconds or minutes of lost GPS tracks is not appropriate. Instead a snapshot of the current position of all tracked entities is preferable. Similarly in a multimedia stream or VOIP call, just the key missing information should be replayed. Not only does full retransmission of all missing packets introduce delays during retransmission, but also the bandwidth consumed during retransmission is wasted. Note that each application may have different semantics. Therefore determining what data is appropriate to deliver for a given application in order to recover from a temporary disconnection requires knowledge of the application semantics. In previous work, we proposed Heterogeneous Intelligent Filtering (HIF), a technology that intelligently filters and transforms data to match network capacity and end-user capability. In HIF, extraneous information not needed by the end-user is filtered by HIF agents. In this work2, we show how the HIF concept can be extended to the data and information persisted during disruptions. The information stored for later retransmission to previously disconnected receivers may consist of application state snapshots, summaries or transcripts of the missing data, or just those portions of a media stream judged to be relevant to the user at a given point in time. We describe our work on application aware persistence for real-time multimedia and tactical situational awareness applications. The middleware infrastructure of the HIF systems reported on in previous work is extended to support application aware persistence. We analyze the performance benefit of application persistence on a tactical situational awareness scenario. Application persistence is particularly useful when the end-device may be limited in its processing power and ability to render data, since it reduces the demands on the client application. Accordingly, we also illustrate the usefulness of our application aware persistence middleware in supporting multimedia and situational awareness applications implemented on handheld Android devices.