Diverse QOS requirements in Integrated Broadband Networks : Roles of Priorities and In-Call Negotiations

Deborah Estrin Computer Science Dept. University of Southern California estrin@usc.edu (213)740-4524 (213)740-7285 (fax) My research interest is quality of service support in LARGE high speed networks, in particular: routing, interaction of routing and resource reservation/management, user incentives/charging model, and simulation techniques, all for very large internetworks. Each of these is described, briefly, below. Several of the papers listed provide more detailed discussions. 1. The increasing speed of underlying transmission media is enabling new, and perhaps more importantly, integrated applications. However, because of the size and decentralized nature of the global internet, we cannot assume that the network will be homogeneously capable of all new services. Therefore, routing must identify routes with specific type-of-service (TOS) capabilities. Moreover, we suggest that TOS-routing for performance-sensitive applications (e.g., real-time video-conference, real-time interaction with high-bandwidth sensor or experimental data) should be sensitive to significant load changes, as well as to topology changes. At the same time, global scale means that we must minimize reliance on global consistency of routing databases and global distribution of dynamic routing information. As a basis for supporting application-specific routing we propose the use of source routing and a limited distribution mechanisms called Reverse Path Update. Source routing allows a wide range of routes to be used in the network without burdening intermediate nodes with agreement on, or precomputation of, all possible routing choices. Source routing also allows us to route without loops even in cases where different network nodes may maintain different (inconsistent) routing databases and criteria. We are exploring a technique called Reverse Path Update (RPU). RPU supports real time requirements, while reducing routing overhead by exploiting locality. It achieves this by distributing status and dynamic-load information only along ‘‘active routes’’. In other words, when an inter-domain connection experiences a change in state (i.e., fails, resumes service, passes a load-threshold), the domain that detects the change distributes the information only to the sources of routes that currently pass through the domain. This does not mean that information is distributed only to sources that pass through the affected physical router; in order to exploit route locality, information is distributed to active routes that pass through any of the border routers belonging to that same domain. The load information distributed indicates long-term load as opposed to short-lived fluctuations. We have documented preliminary results from simulation (see paper below), but many more research questions remain unaddressed, in particular, selection algorithms, load measures, and interaction with resource reservation and flow control mechanisms. (Joint work with L. Breslau (PhD student) and L. Zhang (Xerox)) 2. To support real-time traffic across a heterogeneous internet, resource management and routing must interact in a complementary manner. An obvious example is that stated above: if a particular flow can not be accommodated along a particular path due to insufficient resources (i.e. excessive traffic) then it may be desirable to provide an alternative route. Another routing issue is multicast. We must adapt current IP multicast approaches to interact with resourcereservation mechanism (i.e., what happens when part of a multicast tree does not have sufficient resources to provide the service guarantees required by a flow), and to scale more efficiently (i.e., allow senders to install/define small multicast groups/routes on demand when the entire group is known and is very small compared to the global internet). Beyond the problem of identifying a usable route, is the interaction between resource reservation/management techniques and adaptive routing. Some networks may require explicit flow setup and resource reservation to precede data transmission, and may install flow or flowtype specific state to provide service guarantees. Other networks, ALONG THE SAME sourcedestination path, may be very resource rich with respect to individual flows and may not require flow-specific setup or state. We need an internet protocol that allows individual networks along a path to make independent service guarantees in the presence of diverse network resources,