Proceedings of the 8th IEEE Symposium on Computers and Communications - ISCC'2003, Antalya, Turkey, June 2003 June 2003, pp.1435-1440., IEEE Computer Society Press (unformatted version) Symbiotic Streaming of Elastic Traffic on Interactive Transport 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|rzaghal@cs.kent.edu Abstract Interactivity in transport protocol can greatly benefit transport friendly applications. We envision if a transport mechanism, which is interactive and can provide event notification about network state to the subscriber of its communication service, than a wide range of solution to many of today’s hard network problem can be instituted. Recently we have implemented this concept system as a new TCP kernel on FreeBSD called TCP Interactive and a novel symbiotic MPEG-2 full logic transcoder. In this paper we share the results of the TCP Interactive performance experiment and show potential dramatic improvement in time-bounded video delivery. Index terms-- netcentric applications, TCP interactive, transcoding, temporal QoS. 1. Introduction Recently, TCP-friendliness has been suggested applications for congestion control [12]. However, a major problem in attaining any such friendliness is that the network itself extends very little help to the applications to encourage friendliness. A particular difficult problem is the congestion control for time- sensitive multimedia traffic. Most of the mechanisms for congestion control those have been proposed to date are based on delaying traffic at various network points. The more classical schemes depend on numerous variants of packet dropping in network, prioritization (graceful delay in router buffer) admission control (delaying at network egress points), etc. However, a key aspect of a vast majority of these schemes is that they introduce time distortion in the transport pathway. Though time distortion does no harm to time insensitive traffic, but they work completely against the application if the traffic is time sensitive such as multimedia streaming or control data. In this context, this paper presents a new approach. The suggestion does not require any new protocol per se but suggests adding interactivity with subscriber layer within current transport protocols. We then demonstrate effective solution to the hard problem of congestion control for time sensitive traffic based on this principle of protocol interactivity. The general principle we follow is simple and intuitive. It seems an effective delay conformant solution for time sensitive traffic may be built if the original data volume can be reduced by its originator-- the application. However, a key element in any such scheme is that the application must be notified. Unfortunately, today’s transport protocols do not support any interactivity with applications. It seems such non- interactivity has been inherited from the early days of networking interface research, when the applications were simple 1 . In this paper we will show that transport interactivity can bring major benefit to high performance and demanding applications. The particular scheme we propose suggest a new paradigm of TCP friendliness based on enhanced network and its subscriber layer interactivity and has the following novel aspects: § First, we suggest an active and direct notification mechanism by the underlying transport protocol, rather than using indirect end-to-end feedback tools. § To demonstrate the efficacy of the principle, we have designed a corresponding video rate transcoder system that works in symbiosis with the network. This transcoder actively participates in a custom symbiotic exponential-back-off and additive-increase like scheme in application layer with deep application level knowledge. (This is also one of the first to our knowledge) resulting in much more effective joint quality/delay sensitive communication. § The resulting scheme is similar in spirit to the TCP friendly approaches. However, there is a fundamental difference in how it is done. We expect network (or system) layers to remain as simple as possible. The means and techniques for rate reduction remain with the producer application. The responsibility of the network 1 Interestingly the original TCP proposal (RFC007, RFC793) did call for asynchronous interlayer interactivity but it was never implemented. Need for some forms of callback facility have been marginally cited in recent literature, unfortunately without any technical depth. A survey is available at [13].