A preplanned path-protection scheme with sufficient spare bandwidth is appropriate for real-time fault restoration in multiprotocol label switching (MPLS) networks. In this case, however, it is important to reduce the amount of spare bandwidth to prevent degradation of network efficiency. A distributed label switched path (D-LSP) scheme is proposed to reduce the amount of spare bandwidth required for protecting against network faults in MPLS networks. The main idea of the proposed D-LSP scheme is to partition traffic into multiple LSPs, each of which is established on a distinct link-disjoint route between each pair of end nodes. The D-LSP scheme is evaluated in terms of the reduction ratio of total network cost in comparison with the conventional LSP scheme. Traffic partitioning in the D-LSP scheme can decrease the statistical multiplexing gain (SMG) obtained by aggregating IP packet flows into an LSP. The tradeoff between spare bandwidth reduction and degradation of SMG due to traffic partitioning is also investigated. The numerical results show that the proposed D-LSP scheme yields the network cost-reduction ratio (NCRR) of at least 29%, 27%, and 15% for the networks where average node degrees are 4.6, 4.4, and 3.2, respectively. The D-LSP scheme shows the similar performance of NCRR in both Markovian traffic and self-similar traffic environments.

Spare bandwidth is required for recovering the network service from network faults. However, it degrades the efficiency of network utilization. Spare bandwidth demand can be reduced significantly by letting spare bandwidth be shared among several network services. Spare bandwidth reserved on a network element can be shared by a set of network services for a network fault if they are not simultaneously affected by the network fault. A new, and more practical spare bandwidth sharing scheme, which is based on the network reliability, is proposed in this paper. In the proposed scheme, multiple link failures are allowed with a given link failure rate, and a reasonable restoration level of near 100%; while in the conventional scheme, only a single link failure, and 100% restoration level are considered. To develop the spare bandwidth sharing scheme, we first investigate the framework for evaluating the reliability of path-based network services, and then we explain the proposed spare bandwidth sharing scheme with decision parameters such as lifetime of the path, restoration level, and the maximum number of working paths which can be protected by a backup link. Simulation results show that the proposed spare bandwidth sharing scheme requires a smaller amount of spare bandwidth than the conventional scheme.

We propose a new label switched path (LSP) mechanism, called the distributed LSP (D-LSP) mechanism, for multiprotocol label switching (MPLS) networks. In the proposed D-LSP mechanism, traffic belonging to a forwarding equivalent class (FEC) is partitioned into several sub-classes and is distributed to the corresponding sub-LSPs, each of which is set up on a different node-disjoint route connecting the traffic source and destination node pair. To protect service, a D-LSP reserves the spare bandwidth of which the amount is equal to the amount of the working bandwidth of one sub-LSP and thus the total amount of spare bandwidth in a network can be reduced. The traffic partitioning of the D-LSP mechanism may decrease the statistical multiplexing gain obtained by aggregating IP packet flows into an LSP, as compared to that of the conventional MPLS networks. Sub-LSP and backup LSP assignment algorithms are proposed and the effect of D-LSP mechanism on the required amount of spare bandwidth is evaluated by numeric analysis and is compared with a conventional LSP mechanism. The results show that there is a trade-off between the reduction of spare bandwidth and the degradation of statistical multiplexing gain.