Live video scheduling in differentiated services internet

Abstract

Future Internet services will be judged to a large degree by the efficient deployment of real-time applications. The issue of utilisation is especially central for the economic introduction of live video in packet switching networks; at the same time, the bandwidth requirement is high and the QoS guarantees are stringent. With static allocations based on arbitrary policy enforcement and scheduling mechanisms coupled to them this cannot be achieved. Instead, we propose a fast intra-domain bandwidth management approach that can help in rational policy decision making. It is based on inexpensive video smoothing and scheduling based either on adaptive tracking of aggregate video queue occupancy, or adaptive prediction of aggregate video arrivals. In this way, a general measurement-based utilisation theory for live video can be derived that ensures respectively constant maximum allocation utilisation guarantees or less-than-maximum allocation utilisation target following. QoS guarantees take the form of a constant maximum per-hop delay that is solely dependent on tracker/identifier sampling interval selection. Overall, we show how to provide support for live video under differentiated services.