Mohand Yazid Saidi

Improved dual-forest for multicast protection

To provide fault-tolerance for multicast connections, different techniques of protection are developed. These techniques can be classed into reactive and pro-active approaches. Reactive approaches can have long recovery latency which is undesirable for many types of applications such as the real time ones. In this paper, we focus on the multicast pro-active fault-tolerance schemes. One of the promising protection techniques is the dual-tree protection which is efficient to cope with single link failure but which cannot deal with node failures suitably. In this paper, we present an improved solution using a dual-forest for multicast protection. Our proposition provides three improvements to traditional dual-tree protection. The first one concerns the capability to bypass both single link and node failures in a suitable and quick manner. The second increases the level of protection with the use of a forest as backup instead of a tree. The last permits the cost optimization of the dual-forest and of the delivery tree after recovery. Simulation experiments show that the improved dual-forest scheme has better protection rate and causes less tree cost increase after recovery than the path-protection scheme

Targeted Distribution of Resource Allocation for Backup LSP Computation

Under the hypothesis of single failures in the network, some backup paths cannot be active at the same time because they protect against the failure of different components. Hence, share the bandwidth between such backup paths is central to optimize the bandwidth allocated in the network and to decrease the bandwidth wasting. In this paper, we propose a novel algorithm, based on targeted distribution of resource allocation (TDRA), to compute the backup label switched paths (LSPs) in a distributed multiprotocol label switching (MPLS) environment. Our algorithm is scalable, efficient and capable to protect against the three types of failure risk: node, link and shared risk link group (SRLG). Indeed, the TDRA algorithm decreases the quantity of information (resource or bandwidth allocation) transmitted in the network with the selection of nodes to be advertised (with the selection of recipient nodes). Furthermore, bandwidth availability is increased by sharing bandwidth between backup LSPs as long as possible. Simulations show that the ratio of rejected backup LSPs obtained with the transmission of a small quantity of information in the network is low.

A Distributed Bandwidth Sharing Heuristic for Backup LSP Computation

With the advent of MPLS, the restoration times of communications is decreased down to 50 ms by the use of preconfigured backup LSPs. To ensure there are enough resources after a failure, the backup LSPs must reserve the resources they need beforehand. However and contrarily to the primary LSPs which really use their resources, the backup LSPs do not use them until a failure of the protected component occurs. Hence, to optimize and maximize resource availability in the network, backup LSPs may share their resource reservation. Indeed, under the hypothesis of single failures in the network, some backup paths are not active at the same time since they protect against the failure of different components. In this article, we propose an efficient Distributed Bandwidth Sharing (DBS) heuristic capable to protect the primary LSPs against all types of failure risks (link, node and SRLG risks) with the transmission of a very small amount of bandwidth information. Our technique is completely distributed; it balances the computations on the different nodes of the topology and is easy to be deployed. Simulations show that with the transmission of a small vector of bandwidth information per link, the rate of rejected backup LSPs is low and close to the ideal.

Distributed PLR-based backup path computation in MPLS networks

In this article, we provide mechanisms enabling the backup path computation to be performed on-line and locally by the Points of Local Repair (PLRs), in the context of the MPLS-TE fast reroute. To achieve a high degree of bandwidth sharing, the Backup Path Computation Entities (BPCEs), running on PLRs, require the knowledge and maintenance of a great quantity of bandwidth information (non aggregated link information or per path information) which is undesirable in distributed environments. To get around this problem, we propose a distributed PLR (Point of Local Repair)-based heuristic (DPLRH) which aggregates and noticeably decreases the size of the bandwidth information advertised in the network while maintaining the bandwidth sharing high. DPLRH is scalable, easy to be deployed and balances equitably computations on the network routers. Simulations show that with the transmission of a small quantity of aggregated information per link, the ratio of rejected backup paths is low and close to the ideal.

Hybrid heuristic for Capacitated Network Design Problem

This paper deals with the NP-hard Capacitated Network Design Problem (CNDP) with modular link capacity structure. We aim to optimize the underlying network, the allocation facilities and the transportation costs. It is assumed that for connecting two nodes, a fixed construction link cost is occurred or a specific unit transportation cost. There are several types of module facilities to allocate on links. In this paper, we propose a general model that combines design, allocation and routing problems. An hybrid approach is proposed which uses a greedy constructive and an iterative local search heuristics. Our proposition is tested on the Survivable fixed telecommunication Network Design Library (SNDlib) and a comparison of its performances against the best solution available in the literature is done. Computational tests show that our proposition is very efficient.

PLR-based heuristic for backup path computation in MPLS networks

To ensure service continuity in networks, local protection pre-configuring the backup paths is preferred to global protection. Under the practical hypothesis of single physical failures in the network, the backup paths which protect against different logical failure risks (node, link and shared risk link group (SRLG)) cannot be active at the same time. Thus, sharing bandwidth between such backup paths is crucial to increase the bandwidth availability. In this article, we focus on the optimal on-line distributed computation of the bandwidth-guaranteed backup paths in MPLS networks. As the requests for connection establishment and release arrive dynamically without knowledge of future arrivals, we choose to use the on-line mode to avoid LSP reconfigurations. We also selected a distributed computation to offer scalability and decrease the LSP setup time. Finally, the optimization of bandwidth utilization can be achieved thanks to the flexibility of the path choice offered by MPLS and to the bandwidth sharing. For a good bandwidth sharing, the backup path computation entities (BPCEs) require the knowledge and maintenance of a great quantity of bandwidth information (e.g. non aggregated link information or per path information) which is undesirable in distributed environments. To get around this problem, we propose here a PLR (point of local repair)-based heuristic (PLRH) which aggregates and noticeably decreases the size of the bandwidth information advertised in the network while offering a high bandwidth sharing. PLRH permits an efficient computation of backup paths. It is scalable, easy to be deployed and balances equitably computations on the network nodes. Simulations show that with the transmission of a small quantity of aggregated information per link, the ratio of rejected backup paths is low and close to the optimum.

Using Shared Risk Link Groups to enhance backup path computation

To cope quickly with all types of failure risks (link, node and Shared Risk Link Group (SRLG)), each router detecting a failure on an outgoing interface activates locally all the backup paths protecting the primary paths which traverse the failed interface. With the observation that upon a SRLG failure, some active backup paths are inoperative and do not really participate to the recovery (since they do not receive any traffic flow), we propose a new algorithm (SRLG structure exploitation algorithm or SSEA) exploiting the SRLG structures to enhance the admission control and improve the protection rate. With our algorithm, more flexibility is provided for the backup path selection since a backup path which protects against the failure of a link belonging to a SRLG does not systematically bypass all the links of that SRLG. Moreover, our algorithm permits to save more bandwidth because it does not allocate the bandwidth for the inoperative backup paths even if they are activated. Simulations show that our algorithm SSEA decreases the ratio of rejected backup paths and, it reduces in distributed environments the average number of messages sent to manage the bandwidth information necessary for the backup path computation.

Multi-domain virtual network embedding with coordinated link mapping

Network Virtualization, which allows the coexistence of various logical networks on shared physical infrastructure, has become popular in recent years. The optimal mapping of virtual resource to physical resource is a major issue in network virtualization. This problem, called virtual network embedding (VNE), has been well explored in the context of one physical domain, which is in practice operated by a single infrastructure provider (InP). However, the needs of virtual network (VN) is rapidly growing, and quite a number of VNs have to be established across multi-domain. For multi-domain VNE, infrastructure providers can no longer just solve their own single domain VNE problem, but have to cooperate to build the whole VN. Therefore, new challenge arises for the multi-domain VNE, compared to traditional single domain VNE. The existing investigations on this problem mainly focus on decomposing a VN to sub VN for each domain, but little attention has been paid to the joint relation between intra-domain and inter-domain (peering) links. In this paper, we propose a multi-domain link mapping method which combines the intra and peering link mapping so as to optimize the overall resource utilization. Our approach is easy to be deployed since it is based on current Internet architecture. Evaluation shows that our approach brings improvements related to existing methods.

Resource saving: Which resource sharing strategy to protect primary shortest paths?

Two strategies of resource sharing are proposed in literature to provide protection while saving resources: (1) restrained sharing which applies the resource sharing to the backup paths only and (2) global sharing which extends the resource sharing to the primary and backup paths. In this paper, we compared the two strategies of resource sharing when the primary paths correspond to the shortest ones according to a strictly positive and static metric. Even when the amount of resources that can be shared between the primary and the backup paths is unbounded, we proved that the maximum number of backup paths is still bounded. Besides, our simulations showed that the resource sharing between the primary and backup paths has very slight impact on the backup path rejection, i.e. the two strategies of resource sharing have very close performances.

Two level evolutionary algorithm for Capacitated Network Design Problem

In this paper, we deal with the Capacitated Network Design Problem (CNDP) with modular link capacities to design minimum cost network while satisfying the flow demands. We propose a two level Genetic Algorithm (GA) based model that can deal with several variations of CNDP. Our proposition defines a new encoding scheme to treat the modular case. Extensive simulation results on Atlanta, France and Germany network instances show that the proposed algorithm is much more efficient than the Iterative Local Search algorithm.