3. PROPOSED ALGORITHMThis section describes the proposed algorithm named Traffic Engineering routing Algorithm with Routing Data (TEARD). In order to make the algorithm adaptive to routing demands, link weights are calculated from not only network topology and residual bandwidths but also from routing data.Firstly, ingress egress pairs are considered. For each ie pair, link criticality is determined as the occurrence rate of the link in all paths of the pair. For instance, if a pair ie has 5 paths and a link l appears in 3 of them, then the criticality of link l for the pair ie is critie(l)=3/5. This criticality is calculated in the offline phase because it is determined by the network topology only. Furthermore, in the online phase, the link criticality for each pair is multiplied by probability of the pair being requested. This modification makes the critical values dynamically adapt to actual routing requests. For example, when link l appears in many paths of pair ie, the critical value critie(l) is high and the algorithm will try to avoid this link. However, if ie is infrequently requested, such avoidance may have negative effect on the overall routing performance. In this case, the multiplication by low probability of ie helps reducing critie(l). The link criticality for ie pairs is calculated as follows:
đang được dịch, vui lòng đợi..