Technical Report 2003-09-01 Internetworking and Media Communications Research Laboratories Department of Computer Science, Kent State University http://medianet.kent.edu/technicalreports.html Dynamic Transport Enhancement for Time Elastic Traffic with Transientware Javed I. Khan, Raid Zaghal, and Qiong Gu Networking and Media Communications Research Laboratories Department of Computer Science, Kent State University 233 MSB, Kent, OH 44242 javed|qgu|rzaghal@kent.edu 17 April 2003 Abstract— The paper presents the concept of transientware—a mechanism by which sensitive applications can launch application level targeted and dynamically adaptive enhancements of transport service. The proposed transientware can be used with current transport protocols with interactivity enhancement. As a proof of concept we have recently designed and implemented a functionally enriched transientware enabled TCP on FreeBSD Unix. We then show the performance of a TCP friendly adaptive MPEG-2 video transcoder, which can directly interact with this transport and adjust its outgoing video rate to satisfy temporal quality constraint of the stream via a dynamic rate adaptive scheme. We report potential dramatic improvement in time-bounded video delivery of elastic traffic from live experiment. Index terms-- netcentric applications, TCP interactive, transcoding, temporal QoS. 1 INTRODUCTION 1.1 Transient-ware for Dynamic Adaptation: The interface between and application and network is emerging as a critical research problem in internet base advanced applications research. In this paper we discuss a method where demanding applications themselves can optionally append lightweight transient coupler processes called “T- ware”, which can respond to dynamic network events. It provides application aware smart transport enhancements on the transport services provided by network layer. The T-ware solution is particularly attractive for emerging areas with non-traditional transport needs such as space networking, high performance grid applications, and instrument networking. The approach does not require a static middleware in the traditional sense in the data path between application and network. Rather, these are optional and transient and only invoked conditionally when required. Nor it requires any new or any complex change in conventional transport mechanisms. Also, the transport enhancers can be very effective and efficient as they can implement domain knowledge enriched fully application aware solutions targeted to meet the applications specific need. These ‘T-wares’ does not interfere with the Internet’s network dynamics. Rather it enables applications to be more network-aware for its own good. The ‘T-ware’s enhancers can be used to implement smart solutions to many network transport deficiencies of current transport systems by application programmer themselves (or third party enhancement experts). As a proof of concept we have recently implemented the ‘T-ware’ provision on a FreeBSD Unix. In this paper we present how the T-ware helps an application in its effort of dynamic adaptation in the face of dynamic network congestion while sending time sensitive elastic traffic. The solution has been recently tested for high fidelity video streaming over worldwide network over ABONE with dramatic improvement in time bound quality [KhZh03]. This paper presents the conceptual framework for the T-ware. 1.2 Example Case: Congestion Response for Time Sensitive Elastic Traffic The problem to which we show a smart solution via T-ware is quite challenging. Coping with the dynamically varying characteristics is an open problem in network applications. In asymmetric networks, such as wireless, very long haul networks, or mobile or power aware device