Multilayer Recovery Mechanism for IP over WDM Networks

X Liu,QJ Zeng,Y Wang
DOI: https://doi.org/10.1117/12.523839
2004-01-01
Abstract:In this article, we mainly focus on the integrated dynamic survivability in IP/MPLS/WDM networks under the peer-modeled architecture. We systematically investigated various integrated dynamic protection and restoration schemes combining knowledge of resource and topology information in both the IP/MPLS and WDM layers. To allocate the restorable bandwidth-guaranteed paths, we propose an integrated routing algorithm: Load-Balanced Physical Resource Minimum Hop (LBPRMH) integrated routing algorithm. In this algorithm, we present a new link cost function used in Dijkstra's algorithm considering both physical resource and load balancing factors and employ a parameter (relevant to traffic intensity) to control the influence of two kind of information. When network load is low (traffic intensity is small), the main factor in the proposed link cost function is physical resource, which leads to high resource utilization. When network load is high, the idea of load balance becomes more important in routing, which avoids connection requests conflict and increases the efficiency of resource usage. Then extensive simulations are done to analyze and compare the dynamic survival performance of various integrated survival strategies. Resource utilization, protection efficiency, restoration efficiency are employed as metrics to evaluate the survivable performance. We demonstrate that the proposed integrated routing algorithm performs better than some existing approaches in terms of the above three metrics. The simulation results show the high resource utilization in optical layer recovery with wavelength granularity and the high restoration flexibility in IP/MPLS layer recovery with finer granularity. If the recovery action is implemented first in lower layer and then higher layer if the first action failed, the integrated protection and restoration schemes can combine the advantages of both layers and perform better than recovery schemes in either single layer.
What problem does this paper attempt to address?