Green optical networks with connection availability guarantee

Xiaoning Zhang; Shufang Tan

doi:10.1364/OE.23.017456

1. Introduction

In recent years, the development of ICT (Information and Communications Technologies) requires enormous energy consumption, resulting in the increasingly severe global energy crisis and the worsening greenhouse effect [1]. With the explosive growth of the scale of networks and the rapid advances of the Internet technology, the power consumption of networks will keep increasing at very fast speed [2].

In backbone networks, the energy consumption of optical network equipment such as transponders, optical amplifiers and OXC (Optical Cross Connect), plays a significant part of electricity cost. For the survivable optical networks, many redundant resources would exist in the network so that the high power consumption would be obtained. In this case, the fundamental method, by which redundant resources can be switched off or put into a low-energy mode as many as possible, can effectively aim at saving energy in optical networks. The low-energy mode defined as sleep mode is employed only for backup resources in order to save energy [3,4].

However, most of the studies only consider the problem of energy efficiency in optical networks, while few studies are discussed for both energy efficiency and connection availability, even their mutual relationship. In fact, connection availability is defined in the SLA (Service Level Agreement) along with the service reliability. It’s unwise to consider energy-efficient optical networks without connection availability guarantee. Thus, we study green optical networks with connection availability guarantee in this paper. We introduce a power consumption model for green optical networks design. Our idea is to evaluate the power consumption of components including source node, destination node, links and intermediate nodes. This power consumption model can describe the basic condition in respect of energy consumption and help to increase the potentialities of energy-efficiency. We then propose a new design method called Availability-Guaranteed Energy-Efficient Design (AGEED). In AGEED, we employ dedicated path protection as backup capacity in order to improve the availability and reduce the disabled risk. And then we can make some resources be switched to sleep mode, thus obtain an energy-saving effect if there is no heavy traffic. In order to guarantee the optimal energy-efficient result, AGEED is mainly based on iterative routing by modifying link weight in each iterative loop. The ultimate purpose is to make more links into sleep mode and only have a slight impact on connection availability.

The rest of the paper is organized as follows. Section II briefs the related work of green WDM optical networks. Section III provides a network model and a power consumption model for the green optical networks. Section IV presents the proposed scheme AGEED in detail, including sleep mode and iterative method. Section V presents and discusses the simulation results. Section VI concludes the paper.

2. Related work

With the explosive growth of the scale of networks, recent studies of designing or operating networks have taken into account energy consumption. So the employment of energy efficiency becomes a growing trend in this field. In reference to energy-efficient optical networks, there can be divided into two kinds approximately, the single-layer and the multi-layer. For the single-layer energy-efficient optical networks, low power mode or nearly zero-power mode has been considered as an effective strategy to redundant resources. A. Muhammad et al. [3] advocate a sleep mode for the redundant resources provided for protection purpose only, and present an ILP formulation to guarantee optimal results for energy-efficiency with dedicated-path protection. Similarly, C. Cavdar et al. [4] employ the sleep mode for the backup resources to achieve an energy-efficient, shared backup protected network, and propose an ILP formulation for this problem. A. Coiro et al. [5] present a power consumption model for the optical devices, and an ILP formulation to solve the power minimization problem. They also propose several link-ordering criteria, and apply them to an optical links switch-off algorithm for energy minimization.

As above, energy-efficiency design for optical networks has not been considered completely, especially with the service reliability, one of the customers’ main concerns. Actually, connection availability is representative of service reliability in the SLA [6]. In order to study the availability of green optical networks, B. Kantarci et al. [7] present a network design scheme named Power-Aware Reliable Design (PARD). PARD can lead to a critical decrease in energy consumption; while that only cause a slight decrease in availability levels for the traffic. Likewise, we extend the energy-efficient work to include connection availability guarantee. But unlike PARD, our idea is employing a sleep mode and an iteratively routing mechanism for energy efficiency with high connection availability.

3. Problem statement

In this paper, our objective is to solve the problem of routing and wavelength assignment at minimum power consumption with dedicated path protection. If the network loads are heavy, then it is difficult to merge active links for energy-efficiency. So we use the energy-efficient scheme when the network loads are not heavy. We plan to find the solution under the network model and the power consumption model described as follows.

3.1 Network model

This section introduces the considered optical network scenario for the power consumption minimization problem with dedicated backup path protection in case of single-link failures. By routing in the optical networks, both single-layer energy-efficient approach and multi-layer energy-efficient approach can be used to minimize the total power consumption. In our study, single-layer energy-efficient approach would be adopted to analyze.

The physical network topology can be described as G (V, E, W), where V is the set of physical nodes, E is the set of physical links and W is the set of wavelengths per physical link.

Inputs and notations are introduced as follows.

D A list of demands and |D| is the number of demands.
Bd_k The bandwidth of demand k.
(i, j) The physical link between node i and node j.
(s, d) The connection from source node s to destination node d.
C_ij The available capacity of physical link (i, j).
C_ij^t The available capacity of physical link (i, j)after the t iteration loop.
WW_ij^t The weight of link (i, j) for the t iteration loop while routing a working path.
WDB_ij^t The weight of link (i, j) for the t iteration loop while routing a dedicated backup path.
NB_ij^t Number of backup wavelengths used on link (i, j) after the t iteration loop.
NW_ij^t Number of working wavelengths used on link (i, j) after the t iteration loop.
WP_k A set of the physical links of the working path for demand k.
BP_k A set of the physical links for the backup path of demand k.
d_mn The distance between node m and node n.
d₀ The basic distance between two neighboring optical amplifiers.
a_ij The availability of link (i, j).
A_k The availability of demand k.
t_OS^k The number of optical switching nodes for the path of demand k.
h^k The hop number for the path of demand k.

3.2 Power consumption model

A connection could consist of serial operations, so that the total power consumption can be calculated by operational power [6]. Thus, we evaluate the power consumption of components including source node, destination node, links and intermediate nodes. The power consumption of each component depending on operations is illustrated in Fig. 1. Each node is equipped with an OXC (Optical Cross Connect). When the traffic is from source node A to destination node C, there should be several operations divided into four parts as follows to deal with the traffic.

Fig. 1 Operations of a path routed on an optical network.

Download Full Size | PDF

• Source node A: optically switching (OS) in the OXC to an outgoing port ->transmitting via an optical transponder (TX).
• The fiber link: the signal is amplified by optical amplifiers (AM).
• Intermediate node B: optically switching (OS) by the OXC.
• Destination node C: receiving (RX) ->optically switching (OS) in the OXC to a drop port.

Each of these operations as above will have its own power consumption, such as P_OS, P_TX, P_RX, P_AM. For demand k, the total power consumption (P_total^k) can be divided into four parts shown in Eq. (1), where P_s^k and P_d^k denote the power consumption of source node and the power consumption of destination node. P_in^k denotes the power consumption of intermediate nodes. P_link^k denotes the power consumption of links.

P_{t o t a l}^{k} = P_{s}^{k} + P_{d}^{k} + P_{i n}^{k} + P_{l i n k}^{k}

P_{s}^{k} = P_{O S} + P_{T X}

P_{d}^{k} = P_{O S} + P_{R X}

P_{i n}^{k} = P_{O S} \cdot t_{o s}^{k} = P_{O S} \cdot (h^{k} - 1)

P_{l i n k}^{k} = \sum_{(m, n) \in W P_{k}} P_{m n} = \sum_{(m, n) \in W P_{k}} P_{A M} \cdot (⌈ \frac{d_{m n}}{d_{0}} - 1 ⌉)

As static demand planning is adopted in our study, the fixed traffic will keep the power of source node and the power of destination node in a fixed value for demand k, as shown in Eq. (2) and Eq. (3). Meanwhile, the power of intermediate nodes will be varied with the number of OS (t_os^k) according to the hop (h^k) of a connection, as shown in Eq. (4). P_link^k can be calculated as shown in Eq. (5), while P_mn denotes the power of link (m, n) and it will be varied with the number of optical amplifiers that will be deployed on the link according to the distance d_mn. Therefore, the ultimate goal is to make a critical decrease with P_in^k and P_link^k, while that only cause a slight decrease for connection availability.

4. Availability guaranteed energy-efficient design

In this section, we present the iterative energy-efficient design with connection availability guarantee in optical networks. Simultaneously, we employ a sleep mechanism and an iterative routing mechanism as the main mechanisms in the design.

4.1 Sleep mode of backup resources

In optical networks, optical equipment works in three different operational modes: off, sleep, and active. Mutual transition relationship between these operational modes is shown in Fig. 2, including general situation of the three modes. Although the total power consumed by all the devices depends on their operational modes, we consider that only the devices of links could be in any mode, while the devices of nodes should be in active mode. That’s because a large amount of traffic would lead any pair of nodes to be connected.

Fig. 2 Operations of a path routed on an optical network.

Download Full Size | PDF

Compared to off mode, sleep mode allows quick switching to active mode with negligible power consumption in case of a failure. To implement the sleep mode, some approaches could be adopted. For example, we can turn off most of the circuit and leave critical parts of the circuit on during the sleep mode.

In that case, the entire circuit should be turned off during the off mode. The sleep mode refers to the hardware design, which is not to be described in detail here. As shown in Fig. 2, off mode is employed for idle links with zero power consumption. Once the links accommodate working resources, even only a little, these links must be in active mode. Meanwhile, we assume that links accommodating only backup resources can be switched to sleep mode. Therefore, for all demands, the total power consumption (P_total) can be evaluated by Eq. (6).

\begin{array}{l} P_{t o t a l} = \sum_{k \in D} (P_{s}^{k} + P_{d}^{k} + P_{i n}^{k}) + \sum_{\begin{matrix} (m, n) i n \\ a c t i v e - \mod e \end{matrix}} P_{m n} \\ = \sum_{k \in D} (P_{s}^{k} + P_{d}^{k} + P_{i n}^{k}) + \sum_{\begin{matrix} (m, n) i n \\ a c t i v e - \mod e \end{matrix}} P_{A M} \cdot (⌈ \frac{d_{m n}}{d_{0}} - 1 ⌉ + 2) \end{array}

4.2 AGEED Algorithm

In AGEED, working paths and their corresponding dedicated (link-disjoint) backup paths are routed in optical networks. To reduce power consumption in optical networks, additional optical links are allowed to switch to sleep mode when they accommodate backup resources only and wake-up when the corresponding working paths fail. In this case, we should try to make many links in sleep mode.

An example of network planning is shown in Fig. 3, where two connection requests (1->4, 1->5) are routed. Here, the solid arrow represents a working path and the dotted arrow represents a backup path. Two working paths (1-2-4, 1-3-5) and their corresponding backup paths (1-3-4, 1-2-4-5) are routed in Fig. 3(a), as the initial routing. We can find that links (3, 4) and (4, 5) are redundant while their corresponding working paths are not failed. So links (3, 4) and (4, 5) can be set in sleep mode. However, it’s not enough for energy efficiency, and a better result is expected with the new routing. Thus two working paths (1-2-4, 1-2-4-5) and their corresponding backup paths (1-3-4, 1-3-5) are routed in Fig. 3(b), as the final routing. Notice that three links (1, 3), (3, 4) and (4, 5) can be set in sleep mode, which is more energy-efficient than the initial routing. For this reason, AGEED is implemented to find the expected routing as the final results.

Fig. 3 Working and backup path routing for two connection requests.

Download Full Size | PDF

With the initial routing, we would make few links accommodating backup resources. In such a situation, we need an optimization step to modify link weight and reroute paths. By the optimization step, some backup resources can be aggregated in some links with a slight impact on connection availability. When the optimization step is repeated enough times, we can set many links in sleep mode for energy efficiency. As above, it is our iterative routing mechanism, where each optimization step can be seen as an iterative loop.

The algorithm of the AGEED is described in the following.

Input: a network topology G (V, E, W) and a list of demands D.

Output: static routing and wavelengths assignment at minimum power consumption with dedicated path protection.

Step 1: Pre-processing:
Let k be the index of the demands, and D_k is defined as the demand k.
Define SN_k and DN_k as the source node and the destination node of D_k, respectively.
Define WP_k as a set of links of the work path for D_k.
Define BP_k as a set of links of the backup path for D_k.
Define P_total = {p_t} as a set of fragments p_t that is the power consumption of the t iteration loop.
Define Max_iter as the maximum iteration limit.
Let t be the index of the new iteration loop to be added. Set 0->t.
Step 2: static routing and wavelengths assignment:
Mark all the demands in D as non-Failed. Set 0->k.
2.1) Route a work path for D_k:
2.1.1) Update the link weight for the working path: Based on Eq. (7), the appropriate link weights will be assigned for the working path routed over the logical topology.
2.1.2) Route the working path: Find a shortest working path between SN_k and DN_k by Dijkstra Algorithm. If the working path can be found, update WP_k and go to step 2.1.3. Otherwise mark D_k as Failed, and go to step 2.4.
2.1.3) Assign Resource for the working path: Update the available capacity of each link in WP_k.
2.2) Route a backup path for D_k:
2.2.1) Update the link weight for the backup path: Based on Eq. (8), the appropriate link weights will be assigned for the backup path routed over the logical topology.
2.2.2) Route the backup path: Find the shortest backup path between SN_k and DN_k by Dijkstra Algorithm. If the backup path can be found, update BP_k and go to step 2.2.3. Otherwise mark D_k as Failed, and go to step 2.4.
2.2.3) Assign Resource for the backup path: Update the available capacity of each link in BP_k.
2.3) Evaluate connection availability for Dk: Based on Eq. (9), the connection availability A_k can be evaluated.
2.4) Loop control for static routing and wavelengths assignment:
Set k + 1->k. If k >|D|, go to Step 3. Otherwise go to step 2.1.
Step3: Update and evaluate:
3.1) Update the appropriate mode for all links: For each link in E, if the link accommodates only backup resources, mark it as sleep-mode. If the link is an idle link, mark it as off-mode. Otherwise mark it as active-mode.
3.2) Evaluate the power consumption: According to the mode of links and the power consumption model, evaluate the power consumption and update p_t.
Step 4: Processing for next iteration loop:
4.1) Sort demands: The demands marked as Failed in this iteration loop will be handled with higher priority in next iteration loop by sorting demands. Other demands marked as non-Failed will be sorted after the demands marked as Failed according to the order criterion.
4.2) Record results and release resource: Update all information accordingly, record relevant results and release all resources.
Step 5: Loop control for optimization: Set t + 1->t. If t< = Max_iter, go to step 2. Otherwise return the final solution.

In AGEED algorithm, every iterative loop (from step 2 to step 5) is performed by modifying link weight in the routing algorithm that iteratively tries to switch additional optical links to sleep/off mode as much as possible. By using the proposed algorithm, some paths may not be the shortest and the delay will be increased. However, the transmission time delay is low enough in each optical fiber, so that the end-to-end delay is still able to meet the delay requirement of users. Thus, the time delay wouldn’t be considered here.

At the start, the demands are kept in a list to provision one-by-one. As demand k provisioned, a working path and a backup path should be routed on the physical topology. Equation (7) shows the link weight assignment of each physical link for routing the working path in each iterative loop. Here, the link weight will be a maximum value (∞) while there is not enough available capacity for demand k (where Bd_k>C_ij), and the link weight will take the value 1 at the first iteration loop (where Bd_k≤C_ij and t = 1). ξa, ξb, ξw and ξh respectively denote the coefficients for available capacity, backup wavelengths, working wavelengths and the hop. While under the condition of Bd_k≤C_ij and t>1, the result can be divided into two parts: the right part and the left part. For the high availability level, the right part (P_OS·ξh) is used to avoid too many hops with the working path. Meanwhile, ξa, ξb and ξw are respectively assigned to appropriate values to force the selection of the physical links. In our study, they all equal 0.05. According to the left part, the physical links with lower available capacity, lower number of backup wavelengths and higher number of working wavelengths in the last iteration loop tend to be assigned to the working path for demand k.

W W_{i j}^{t} = {\begin{matrix} P_{i j} \cdot \frac{(N B_{i j}^{t - 1} \cdot ξ_{b} + C_{i j}^{t - 1} \cdot ξ_{a} + 1)}{N W_{i j}^{t - 1} \cdot ξ_{w} + 1} + P_{O S} \cdot ξ_{h} & B d_{k} \leq C_{i j} \land t > 1 \\ 1 & B d_{k} \leq C_{i j} \land t = 1 \\ \infty & B d_{k} > C_{i j} \end{matrix}

After the working path routed, a dedicated backup path is searched on the basis of the weight assignment depending on Eq. (8). Here, it should ensure not only the enough available capacity for routing (where Bd_k≤C_ij), but also the working path and the backup path of the same connection to be link disjoint (where (i, j)∉ WP_k), or else the link weight will be a maximum value (∞). While under the condition of Bd_k≤C_ij and t>1 and (i, j)∉ WP_k, the result can also be divided into two parts: the left part and the right part. As shown in the left part, opposite to the allocation for working path, the backup path prefers to be routed on those links with high number of backup wavelengths and low number of working wavelengths in the last iteration loop. And the right part (P_OS·ξ_h) is still used to limit hops.

W D B_{i j}^{t} = {\begin{matrix} P_{i j} \cdot \frac{N W_{i j}^{t - 1} \cdot ξ_{w} + 1}{N B_{i j}^{t - 1} \cdot ξ_{b} + C_{i j}^{t - 1} \cdot ξ_{a} + 1} + P_{O S} \cdot ξ_{h} & B d_{k} \leq C_{i j} \land t > 1 \land (i, j) \notin W P_{k} \\ 1 & B d_{k} \leq C_{i j} \land t = 1 \land (i, j) \notin W P_{k} \\ \infty & o t h e r w i s e \end{matrix}

As above, the link weight is modified according to the wavelengths assignment in the last iteration loop, except the first iteration loop. With the iteration loops increasing, the results with reference to energy-efficiency and connection availability will be gradually optimized in a compromise.

The connection availability is calculated by the formulation in Eq. (9) while routing a dedicated backup path. Upon condition that the demand’s availability doesn’t satisfy its requirement, the demand is noted as a failure in this iteration loop and handled with high priority in next iteration loop by sorting demands.

A_{k} = \prod_{(i, j) \in W P_{k}} a_{i j} + (1 - \prod_{(i, j) \in W P_{k}} a_{i j}) \prod_{(i, j) \in B P_{k}} a_{i j}

In AGEED, Dijkstra algorithm is used to compute the shortest path between the source node and destination node. The time complexity of Dijkstra algorithm is O(|V|²), where |V| denotes the number of the network nodes. For each demand, AGEED run two times of Dijkstra algorithm to compute working path and backup path. Since there are |D| demands to be deployed in the network and the maximum iteration limit of the algorithm is Max_iter, we conclude the time complexity of AGEED is O(2 Max_iter |D||V|²).

5. Simulation results

For the sake of evaluating the performances of the heuristic algorithm presented above, the network model and a randomly generated traffic demand would be considered as input parameters. Two test networks are employed, including the 14-node 21-link NSFNET network, and the 24-node 43-link USA backbone IP network (USNET, in short), as shown in Fig. 4. The available capacity of each link is 48 wavelengths, and the distance (km) of each link is indicated in the Fig. 4. The availability of each link is assumed to be 0.99. We can calculate connection availability for each demand by Eq. (9), and adopt average availability for all traffic demands as a performance indicator for connection availability analysis.

Fig. 4 Network topology used in our study. (a) NSFNET topology. (b) USANET topology.

Download Full Size | PDF

In AGEED, the iterative routing method is adopted to sleep more links in the network. Thus the value of Max_iter determines the effect of AGEED. The total power consumption vs Max_iter is shown Fig. 5.We can see that for NSFNET when Max_iter equals 10, the minimal energy consumption can be achieved. For USANET when Max_iter equals 12, the minimal energy consumption can be achieved. Thus for different network topology scenarios, Max_iter should be set as different values.

Fig. 5 . (a) Max_iter vs total energy consumption in NSFNET. (b) Max_iter vs total energy consumption in USANET.

Download Full Size | PDF

In order to evaluate the performance of the AGEED algorithm, two methods are implemented as comparisons in the simulation. One is called Selective Switch Sleep/Off of Optical Links (SSSOOL). In SSSOOL described in [5], it starts to route the shortest path by considering all links are active, and then several optimization steps are performed to reduce power consumption. In each optimization step, the algorithm iteratively tries to switch an optical link, which is selected by the fiber ordering criteria, to sleep/off mode. The other method is called Availability Maximizing Design (AMD), which routes the shortest path to guarantee the maximum connection availability. Results in Fig. 6 and Fig. 7 illustrate the comparison.

Fig. 6 Simulation results with NSFNET. (a) total power consumption. (b) power consumption per demand. (c) the average availability. (d) percentage of sleeping links.

Download Full Size | PDF

Fig. 7 Simulation results with USANET. (a) the total power consumption. (b) the power consumption per demand. (c) the average availability. (d) percentage of sleeping links.

Download Full Size | PDF

By performing the AGEED algorithm, the power significantly reduced is mainly based on the optical components described in section III. The data in [2,8] is used for the power consumption of each optical component. The total power consumption is evaluated and shown in Fig. 6(a) and Fig. 7(a). As seen in these two figures, the power consumption of AMD is significantly more than AGEED and SSSOOL. We can see that the red line almost appears to coincide with the green line, thus the energy-saving effect of AGEED performs similar to SSSOOL.

To be fair, we also consider the average power consumption per demand, as shown in Fig. 6(b) and Fig. 7(b). As the traffic load increased, the average power consumption per demand would be decreased. That’s because the optical amplifiers would be shared by multiple traffic, which results in low power consumption per demand. Namely, the more demands traverse, the less power consumption per demand occupies. As seen in the figures, both AGEED and SSSOOL outperform AMD in comparison of the power consumption with dedicated backup path protection.

Figure 6(c) and Fig. 7(c) evaluate the performance in terms of connection availability. AMD leads to the highest availability level for traffic; next are AGEED and SSSOOL. In particular, AGEED guarantees a higher availability level in comparison with SSSOOL. One of our ultimate purposes is to make some links only accommodate backup resources for sleep mode, and results in Fig. 6(d) and Fig. 7(d) illustrate the percentage of sleeping links for energy efficiency. In Fig. 6(d), due to the small network topology for simulation, the percentage of sleeping links is not more than 38.09%. But results in Fig. 7(d) show that the percentage of sleeping links is up to 46.51% with the larger topology. Overall, AGEED cannot only lead to a critical decrease in power consumption, but also guarantee a satisfied connection availability level for the traffic. As seen in the figures, AGEED outperforms AMD in comparison of the power consumption with backup path protection; meanwhile AGEED outperforms SSSOOL in comparison of the connection availability level.

Figure 8 shows demand drop percentage for NSFNET and USANET. In the simulation, we set traffic demand equals 80. When the number of wavelengths increases from 10 to 50, we can see that demand drop percentage clearly decreases. Our proposed AGEED performs best in three algorithms because the weight seting considers the resource constraint in the network. The simulation results from Fig. 8 illustrate AGEED has better resource utilization while saving energy consumption.

Fig. 8 .(a) Demand drop percentage of NSFNET. (b) Demand drop percentage of USANET

Download Full Size | PDF

5. Conclusion

This paper presents a design scheme, namely Availability Guaranteed Energy-Efficient Design (AGEED) for green optical networks with connection availability guarantee. AGEED is mainly based on a sleep mechanism and an iterative routing method. In particular, we studied and discussed connection availability and power consumption as the main performance characteristic. Simulation results show that AGEED is more advantageous to achieve energy-efficient purpose and guarantee satisfied connection availability.

Acknowledgment

This work is supported by National Natural Science Foundation of China (NSFC) under grant No. 61201129.

References and links

1. J. Baliga, K. Hinton, and R. S. Tucker, “Energy consumption of the Internet,” Joint International Conference on Optical Internet and the Australian Conference on Optical Fiber Technology, COINACOFT (IEEE 2007).

2. G. Shen and R. S. Tucker, “Energy-minimized design for IP over WDM networks,” in IEEE/OSA, J. Opt. Commun. Netw. 1(1), 176–186 (2009). [CrossRef]

3. A. Muhammad, P. Monti, I. Cerutti, L. Wosinska, P. Castoldi, and A. Tzanakaki, “Energy-efficient WDM network planning with dedicated protection resources in sleep mode,” in Proc. Globecom (IEEE 2010). [CrossRef]

4. C. Cavdar, F. Buzluca, and L. Wosinska, “Energy-efficient design of survivable WDM networks with shared backup,” Proc. Globecom (IEEE 2010). [CrossRef]

5. A. Coiro, M. Listanti, A. Valenti, and F. Matera, “Reducing power consumption in wavelength routed networks by selective switch off of optical links,” IEEE J. Sel. Top. Quantum Electron. 17(2), 428–436 (2011). [CrossRef]

6. J. Zhang, K. Zhu, H. Zhang, and B. Mukherjee, “A new provisioning framework to provide availability-guaranteed service in WDM mesh networks,” Proc. IEEE Int. Conf. Communications (IEEE 2003). [CrossRef]

7. B. Kantarci and H. T. Mouftah, “Greening the availability design of optical WDM networks,” in IEEE Global Communications Conference 2010 Workshop on Green Communications (IEEE 2010). [CrossRef]

8. I. Cerutti, N. Sambo, and P. Castoldi, “Distributed support of link sleep mode for energy efficient GMPLS networks,” in 2010 36th European Conference and Exhibition on Optical Communication (ECOC), 2010. [CrossRef]

Green optical networks with connection availability guarantee

Abstract

1. Introduction

2. Related work

3. Problem statement

3.1 Network model

3.2 Power consumption model

4. Availability guaranteed energy-efficient design

4.1 Sleep mode of backup resources

4.2 AGEED Algorithm

5. Simulation results

5. Conclusion

Acknowledgment

References and links

Cited By

Figures (8)

Equations (9)

Optics Express