WDM networks. A light-hierarchy is a set of consecutive and directed fiber links occupying the same wavelength, which is rooted from the source and terminated at the destinations. Different from a light-tree, the light-hierarchy structure accepts the cycles introduced by the Cross Pair Switching capability of MI nodes, which enables an MI node to serve several destination nodes on the same wavelength through its different input and output pairs. Light-hierarchy structure overcomes the inherent drawback of the traditional light-tree structure, so that the splitting constraint is relaxed to some extent. This is why it outperforms the light-tree in term of cost. We proved that theoptimal multicast structure for minimizing the wavelength channel cost is not a set of light-trees, but light-hierarchies. ILP formulations are developed and implemented to compute theoptimal light-hierarchies. Numerical results verified that the light-hierarchy structure is the cost optimal solution for all-optical multicast routing with sparse splitting constraint.
Abstract: Most of the existing routing protocols for ad hoc networks are designed to scale innetworks
of a few hundred nodes. They rely on state concerning all links of the network or links on the route between a source and a destination. This may result in poor scaling properties in larger mobile networks or when node mobility is high. Using location information to guide therouting process is one of the most often proposed means to achieve scalability in large mobile networks. However, location- based routing is difficult when there are holes inthe network topology. We propose a novel position- based routing protocol called Proximity Aware Routing for Ad-hoc networks (PARA) to address these issues. PARA selects the next hop of a packet based on 2-hops neighborhood information. We introduce the concept of “proximity discovery”. The knowledge of a node’s 2-hops neighborhood enables the protocol to anticipate concave nodes and helps reduce the risks that therouting protocol will reach a concave node inthe network. Our simulation results show that PARA’s performance is better in sparse networks with little congestion. Moreover, PARA significantly outperforms GPSR for delivery ratio, transmission delay and path length. Our results also indicate that PARA delivers more packets than AODV under the same conditions.
Related work. Inthe context of SRLG, basic network connectivity problems have been proven much more difficult to address than their usual counterparts. For instance, the problem of finding a “SRLG-shortest” st-path that is a path from node s to node t having the minimum number of risks has been proven N P -hard and hard to approximate in gen- eral (see ). However, the problem has been proven to be polynomial in two generic practical cases corresponding to localized failures: when all risks verify the star property  and when risks are of span 1 ; i.e. when a link is affected by at most one risk and links sharing a given risk form a connected component .
I. I NTRODUCTION
Geographic routing –, relying on the knowledge of geographic location information of nodes to make local route decisions, is a promising routing solution to the demand of developing efficient and scalable protocols in multihop wireless ad hoc networks. In recent years, with the significant advances of physical layer transmission techniques, cooper- ative communication for wireless networks has become an active research area due to its ability to create spatial diversity via node cooperation. There have been various cooperative diversity schemes inthe literature –. However, most of them focus on the design of physical-layer cooperative relaying schemes, in which different issues such as signaling strategies, power allocation, relay selection, and bandwidth efficiency are taken into consideration to assess their physical layer performances.
Springer Science+Business Media New York 2015
Abstract There is no doubt that P2P traffic mainly video traffic (e.g. P2P streaming, P2P file sharing, P2P IPTV) increases and will represent a significant percent of the total IP video traffic (80 percent by 2018 of the global IP traffic according forecasts). Peer-to-peer (P2P) is based on some main concepts such as mutualization of resources (e.g. data, programs, service) at Internet scale. It is also considered as one of the most important models able to replace the client-server model (e.g. for media streaming). Nevertheless, one of the fundamental problems of P2P networks is to locate node emplacements or resources and service location. Localisation problem is critical as there is no central server and churn rate can be high in some environments (high dynamicity). Lookup optimization in terms of number of hops or delay is not well considered in existing models, and still represents a real challenge. In this context and according to their specific characteristics and properties, De Bruijn graph based solutions constitute good candidates for lookup optimization. In this paper, we propose a new optimized model for lookup acceleration on P2P networks based on De Bruijn graph. Performance evaluations and simulation results show that our proposed approach is performant, compared to the main existing model.
Mechanisms for explicit path selection are not included in most multicast distribution concepts. With explicit path selection, the sender of a multicast packet can explicitly select the distribution path (usually a tree) of a single multicast packet. This allows a sender selecting individual multi- cast trees for each single packet in order to react on events such as link breaks, node failures, congested links, and group member leaves. We propose that a sender of a multicast packet can select a backup multicast tree instead of the default multicast tree by inserting a fixed size iden- tifier to the multicast packet. A multicast delivery tree is typically established by multicast rout- ing protocols in case of IP multicast and by peer-to-peer protocols in case of application level multicast. Such a multicast delivery tree is then used for the distribution of multicast data. The selected backup multicast tree can then be used to immediately react on link failures without any delay caused by reestablishing a new multicast delivery tree for the new topology. Load balanc- ing can be achieved by using different trees simultaneously and can be applied when a particular link of the default multicast tree becomes congested or for increasing throughput.
As illustrated by Fig. 5, the ad hoc communicating mode can sometimes be faster (the path is shorter) and should be used to speed up therouting process. To determine which mode to use, the node u first asks the value of hc(v) to AP (u). If AP (u) does not have this information, it requests it from other access points inthe wired network. When u retrieves hc(v), it launches a broadcast with a Time-To-Live (TTL) equal to hc(A) + hc(B) − 1 to find a route in pure ad hoc mode (by using DSR for example). If v is not found by using this broadcast, it means that the path between them in ad hoc mode is longer than the one in infrastructure mode (i.e. hc(u, v) > hc(u) + hc(v)). In this case, the infrastructure mode will simply be used. By using this protocol, any two nodes can communicate to each other by knowing their routes and distances to access points.
As noted inthe previous section, EAN are a special case of k-trees, with k = d + 1. There actually exist several results concerning labelings of k-trees that allow to route by shortest paths (see for instance [41, 42]). However, those labelings are too general for our purpose. In particular, they are constructed using the fact that, at each step, a vertex is added to a particular k-clique. However, as mentioned inthe previous section, the notion of step for the usual k-trees is a totally different concept than the one we use for EAN, and as a consequence, their labeling is not of optimal length. Inthe following, we develop a new and original labeling, especially designed for EAN, which is of optimal length and allows to route by shortest paths.
1.4.2 Unicast routing with multiple metrics
Path computing problem for multiple metrics has been widely investigated inthe literature. However the problem is inherently hard and a polynomial time algorithm may not exist. Finding a path with more than one constraint has been proven to be NP-complete . Hence, the algorithms are distinguished by their type of solution, which can be exact, approximate or heuristic. Many works suggest that heuristic solution is the best way to treat them. Jaffe presented two algorithms for the MCP problem with two constraints . The first one is a pseudo-polynomial-time algorithm, the second is a polynomial-time algorithm. It uses an objective function which combines the constraints to create a unique constraint. Then the algorithm finds the shortest path. The closest work is done by Chen and Nahrstedt in , who propose a heuristic algorithm for MCP. The idea is to first reduce the NP-complete problem to a simpler one, which can be solved in polynomial time, and then solve the new problem by using one of the following two algorithms. The first one is an extended Dijkstra’s shortest path algorithm (EDSP). The second one is an extended Bellman-Ford algorithm (EBF). Another algorithm to solve bandwidth-delay-constrained path was proposed by Wang and Crowcroft , which is based on two steps. First, all links that do not respect the bandwidth requirement are eliminated so the output is a feasible graph in term of bandwidth. The second consist in finding the shortest path in term of delay using Dijkstra’s algorithm.
In order to function correctly, however, routing protocols require that each node inthe network is uniquely
identified. Inthe Internet, as we have already seen, this unique name is an IP address. Therefore, there is a
need for a mechanism dynamically assigning unique addresses to nodes in a MANET.
Contrary to traditional networks, the roles of hosts (i.e. nodes that use the network) and routers (i.e. nodes that form and maintain the network) are not clearly separate in a mobile ad hoc environment. Indeed, each node may act in both capacities simultaneously. Another particularity of MANETs is that no assumption can be made regarding preexisting infrastructures. However, classical autoconfiguration mechanisms, such as DHCP , ZeroConf  or similar mechanisms, assume the presence of a server, which can coordinate and assign addresses to hosts inthe local network. Other traditional mechanisms, such as IPv6 Stateless Autoconfiguration , assume that direct communication is available between each interface inthe local network. However, the multi-hop nature of MANETs does not allow the assumption that direct communica- tion between an arbitrary pair of nodes is always possible.
Reservation based scheduling dedicates chan-
nels exclusively for data transmission. For exam- ple, in (Wu, Ke, & Huang, 2007) potential senders use an ALOHA based random MAC scheme to send reservation requests to the central node. As reservation requests may collide and be lost, the reservation process needs an explicit confirmation. The scheduler (using its knowledge of the tuning time and delays) organizes asynchronous data transmissions between senders and destinations. A multicast scheduling algorithm called LBQA (Look Back Queue Access) is proposed. This algo- rithm favors multicast messages which can be sent immediately to all destinations. When there are no more all-receiver messages to transmit and while there are available data channels, the algorithm schedules also partitioned multicast messages (for an available subset of the destinations). The authors state that this scheduling algorithm can also be applied in PONs. The proposed architecture and the scheduling have some drawbacks. The scheduled time slot must allow sufficient time to tune the concerned transmitter and the receivers before data communication can start. This delay limits network performance. A large number of nodes inthe domain can lead to heavy collisions
Fig. 6. Number of rounds generated inthe optimization of cut and flow formulations.
225 nodes: from tenth of seconds to a few minutes for topologies with more than 100 nodes. On one hand, it allows to solve large-scale instances to optimality. On the other hand, the computational time is roughly the same as the formulation with flows : sub-linear inthe network size and linear inthe gateways density. Moreover, one can see in Figure 6 that the number of generated rounds is decreased in comparison to the existing formulation with flows. This is better since the auxiliary program to generate new rounds is an ILP, and is related to the maximum independent set problem which is known to be N P -hard in general graphs. Indeed, if we consider a binary interference model, then a round is an independent set of the conflict graph, i.e. the graph where each node is a possible transmission inthe network, and there exists a link between two of them if the corresponding transmissions are interfering.
Another possibility arises when the time scale over which the fading states vary is much longer than the time that it takes to transmits a unit of information between two nodes. Assuming that the receiver is able to measure the channel state and there is a feedback mechanism for the receiver to send this information back to the transmitter, the transmitter can leverage this information to adjusted the transmitted power based on the present channel state. This approach, however, requires the channel state information at the transmitter. If we assume that such channel knowledge is not available at the transmitter, there is no way that the transmitter can adjust its power to compensate for very bad channel states. The appropriate model for the wireless link in this scenario is the capacity-versus-outage model, see , , . In this model, the instantaneous capacity of a wireless link is treated as a random variable. A link is said to be in outage when the instantaneous capacity supported by the link is less than the transmission rate. The reliability of a link, i.e. the probability of correct reception at the receiver, is modeled as a function of the transmission rate, the transmitted power, the distance between the communicating nodes, and the channel fading state. By adjusting the transmission rate or power, the transmitter can control the probability of successful reception at its intended receiver.
Our motivation stems from the fact that it has been shown that the break-even line length where optical communication lines become more effective than their electrical counterparts is less than 1cm, in terms of speed and power consumption . Therefore, the use of opti- cal interconnections on-board is nowadays justified, and some studies even suggest that on-chip optical intercon- nects will soon be cost-effective . Moreover, the emergence of cutting-edge technologies as Vertical Cav- ity Surface-Emitting Lasers (VCSELs) [15, 31], high sensibility optical transimpedance receivers , beam splitters [17, 18], micro-lenses , and holograms , makes possible the fabrication of complex optical com- munication networks.
in Theorem 3.1. Indeed, let us show with a counterexample that the total length policy is not optimal. In Figure 7 the origin nodes for each packet are labeled with small letters, while their corresponding desti- nation nodes are labeled with capital letters. We have that ℓ A−a = 6, ℓ B−b = 5, ℓ C−c = 5 and ℓ P −p = 4.
One can check that the total number of steps required by the packet that starts at p to reach P is 7 (because it waits 3 steps) and thus this algorithm is not optimal, since 7 > 6 = ℓ max .