Resource-aware provisioning strategies in translucent elastic optical networks

Abstract

Elastic optical networks (EONs) enable a more efficient use of spectrum resources than traditional wavelength-division multiplexing (WDM) networks, mainly by means of a flexible frequency grid and the use of bandwidth variable transponders (BVTs). In WDM networks, regeneration is avoided because it always produces a cost increase and it is only recommended when the transmission length cannot be met. Instead, in EONs, a trade-off exists between regeneration and spectrum use. Indeed, blocking probability can be widely reduced by incorporating regeneration (BVTs in a back-to-back configuration) to shorten the transmission lengths, therefore increasing spectrum efficiency. Provisioning algorithms that take into account this trade-off consider regeneration either as a cost to be minimized (full regeneration capacity) or as a resource to be smartly used (bounded number of predeployed transponders).

In recent years, algorithms that minimize the resource use of either spectrum or BVT have been proposed. These algorithms, however, penalize the overall network performance in terms of blocking probability and cost, since they do not take into account the available resources. In this work, we propose two novel resource-aware algorithms, that take into account all available resources in the provisioning process. We report simulation results showing that using these resource-aware algorithms: (i) when full regeneration capacity is assumed, the regeneration cost can be reduced by more than a 30% compared to the opaque solution and by a 10% compared to the best known algorithms; (ii) for the bounded case, resource-aware strategies can reduce blocking probability by thousands of times compared to the transparent solution and by hundreds of times compared to the best known algorithms. Our results show that resource awareness helps to decrease the blocking probability while increasing the actual network capacity, thanks to a more efficient assignment of both spectrum and BVT resources.

Introduction

Wavelength division multiplexing (WDM) and dense wavelength division multiplexing (DWDM) networks base their operation on a fixed 100/50 GHz grid [1]. Even if these networks can transport large bandwidths they become inefficient in presence of heterogeneous and variable traffic demands. The elastic optical network (EON) architecture [2], [3], proposed almost a decade ago and extensively studied since then [4], enables a more efficient use of spectrum resources by means of a flexible grid structure. In this architecture, demands are provisioned using multiple adjacent 12.5 GHz frequency slots. The amount of slots required by a single demand depends on the demanded bitrate, the transmission length, the modulation format, the baud rate, the number of carriers and the error correction overhead. Bandwidth-variable transponders (BVTs) can adjust the baud rate, the modulation format, the number of carriers and the error correction overhead to transport a demand with a specified bitrate minimizing the required frequency slots. This flexibility enables EONs to establish optical connections with “arbitrary” bandwidths (frequency slots) [4], [5].

On the other hand, regeneration has been widely used in optical networks to reduce the blocking probability when provisioning traffic demands. In traditional WDM networks, regeneration has been mainly proposed to overcome physical impairments that limit the maximum transmission distance [6]. Besides, regeneration can also be used to reduce blocking by providing wavelength conversion capabilities that can tackle spectrum fragmentation issues. However, regeneration in traditional WDM networks has no impact on the spectrum utilization because each lightpath exploits either a 50 GHz or a 100 GHz channel regardless of the path length or assigned wavelength. Instead, regeneration in EONs can also be used to compress the amount of spectrum required for provisioning a demand. To this end, a flexible regenerator composed of back-to-back bandwidth-variable transponders [7], [8], [9] can be used, where each transponder handles the same rate but not necessarily requires the same amount of frequency slots. These regenerators can be configured to support, independently on each carrier, a specific modulation format and baud-rate. The shorter the distance between regeneration points, the higher the modulation format that can be used, leading to an increase in spectral efficiency and spectrum saving. Flexible regeneration enables a trade-off between spectrum and transponder costs, as described in [10], [11].

The use of regeneration and its impact on blocking ratio has been widely studied in the past. Nevertheless, recently it has gained relevance mainly due to the trade-off between regeneration and spectrum use in EONs that we described in the previous paragraph. In recent years, the effort have focused on obtaining provisioning algorithms that take advantage of this trade-off [8], [12], [13], [14], [15], [16]. In this context, regenerators can be considered either as a cost to be minimized or as a resource to be smartly used. In the first case, spectrum is considered as the limiting resource and the use of regeneration to reduce the blocking probability translates to node cost increase. In the second scenario, regenerators are bounded to the amount of transponders that each node is equipped with, at network design phase. The goal becomes to efficiently use available transponders. If spectrum slot lack becomes the main bottleneck when accepting traffic demands, transponders can be used to save spectrum to reduce the demand blocking probability. In operational networks, regeneration capabilities can be increased much faster and smoothly by equipping nodes with more transponders, while increasing spectrum resources is typically slower and may require a huge investment due to fiber cable deployments or renting costs.

In this paper we evaluate different provisioning strategies that tackle both unbounded and bounded regeneration scenarios to illustrate the value of regeneration in EONs. We name these strategies as resource-aware strategies, because they aim at selecting the best combination of available spectrum and transponder resources for establishing a translucent lightpath. Based on our previous contribution [14], originally designed for the full regeneration capacity scenario, we propose a novel strategy, suited to both scenarios. The strategy differs from others proposed in the literature, which are typically agnostic to available resources. We demonstrate that resource awareness helps in decreasing the blocking probability and increases the actual network capacity. thanks to a more efficient assignment of both spectrum and transponder resources. The paper is organized as follows: in Section 2 previous research activities are summarized and compared. In Section 3 we analyze the problem of selecting an optimal provisioning candidate in Elastic Optical Networks. We present the used system model in Section 4. Then we introduce the proposed resource-aware algorithms in Section 5. Finally, in Section 6 we show simulation results and in Section 7 we derive the conclusions.