Preplanned restoration and optimal capacity placement on ATM multicast tree

The ATM multicast Tree (AMT) is the Mbone of video/audio conferencing and o ther multicasting applications in ATM (Asynchronous Transfer Mode) networks [11]. However, real problems such as temporarily moving switches, changing optic fiber connections and/or tangible/intangible failures of ATM network s will cause many service disruptions. Thus we must carefully consider the system's SQOS (Survivable QOS) [7] when we construct the system. A point-to -point self-healing scheme utilizing a conventional pre-planned backup mech anism is proposed to protect the AMT from failure. This scheme uses point-t o-point pre-planned backup Root-to-Leaf Routes (RLR) as the root-to-leaf st ructure of an AMT. Though AMT protection via preplanned backup RLR requires no search time, duplicate paths may cause redundant bandwidth consumption. This paper also proposes a closest-node method, which can locate the minim um-length route structure during the initial design and also rebuild the AM T in the event of a network failure. To enhance the survivability of the sy stem, we introduce two near optimal re-routing algorithms, a most-decent se arch algorithm, and also a predictive-decent search algorithm in order to f ind the minimum lost flow requirement. These near optimal schemes use searc h technique to guide the local optimal lost how to the most-decent lost flo w direction. The predictive way is an especially economical technique to re duce the calculation complexity of lost flow function. For the evaluation o f the feasibility and performance of the new schemes, we simulate AMT resto ration and the simulation results show the closest-node scheme provides sup erior AMT restoration compared to a system with a preplanned point-to-point backup scheme. In addition, the predictive-decent search algorithm is fast er than the most-decent search one.