Why Are You Still Using Shortest Path?

I. INTRODUCTION Live streaming is delay sensitive and their end-to-end allowable delay is an important factor. Allowable delay means the elapsed time in which data should arrive at an end receivers before being played back. Data arriving at end receivers later than the allowable delay are useless and cannot be played back. Therefore, not only must data losses be reduced but also a specific end-to-end data delay must be maintained to ensure the QoS of the application. For live streaming media, routing problems have been an important issue for ensuring the QoS of applications. Generally in unicast routing, the shortest path is often used to minimize the delay from the sender to the receiver. Also, multicast tree configuration depends on the shortest path tree to minimize the delay from the sender to receivers, or the minimum spanning tree to minimize the network resources of the whole tree. As an example shown in Fig. 1, when the traffic conditions of a certain link on the multicast tree become worse and a bottleneck develops during data transmission, a multicast reconfiguration method can avoid the bottleneck link (link B-A) and modify the part of a given multicast tree adaptively to recover and improve the QoS against traffic variations. In multicast tree reconfiguration [1], [2], the shortest delay path is frequently used as an alternate path to ensure a small end-to-end delay. However, the Shortest Path Tree (SPT) reconfiguration is not always appropriate when high-functional nodes are placed on part of nodes on the network. High-functional node is a node with a special capability, such as loss detection, recovery, etc., to improve the QoS of an application, in addition to the functions of normal node. In this situation, the QoS of applications varies depending on the number of high-functional nodes and their locations. However, the shortest delay path selection method does not take into account the existence of these nodes, and thus, it might not be able to select a path with a sufficient number of high-functional nodes. As an example shown in Fig. 1, alternate paths, (C-A) and (D-A) were found. The path (S-C-A-R) is the shortest delay path from the sender to the receiver and has only four high-functional nodes. The other path (S-F-E-D-A-R) has a slightly larger delay than the shortest delay path, but it includes six high-functional nodes. Generally, previous path reconfiguration methods, such as PS-SPT [2], would select the …