Issues and challenges in physical-layer aware optically switched network design and operation

Abstract

The evolution of optical networks should aim at improved cost economics, reduced operations efforts, scalability and adaptation to the future services and application requirements. Considering the evolution trend of optical networks, the core networks of the future will have a translucent and eventually transparent optical structure. The realization of dynamic and fully automated transparent optical core networks is an important task that is required in order to provide cost (CAPEX and OPEX) reduction & performance benefits. This goal has not yet been fully achieved in commercial exploitation due to: a) limited transmission reach and overall performance of transparent optical networks and b) challenges related to the fault localization and isolation in transparent optical networks. These issues have been recognized within the DICONET project. In DICONET framework Physical layer impairments and optical performance are monitored and incorporated in impairment aware lightpath routing algorithms. These algorithms will be integrated into a novel dynamic planning and operation tool that will consider dynamic traffic characteristics, a reconfigurable optical layer, varying physical impairment and component characteristics. The planning and operation tool along with extended control planes will make possible to realize the vision of optical transparency. In this talk we focus on the impairment aware lightpath establishment problem. Two broad techniques are available to solve this problem: the utilization of a quality of transmission estimator based on real-time monitor measurements and analytical transmission models (called hereafter Q-Tool); and through the direct probing of candidate lightpaths. We present recent developments for these two techniques. The first technique, impairment aware routing and wavelength assignment (IA-RWA) using a Q-Tool, has been well investigated especially when lightpaths must be established under a single constraint (e.g., quality of signal transmission must be high enough, or accumulated noise must be below some threshold). Here, we present a multi-constraint IA-RWA algorithm, called “Rahyab”, which uses a single mixed cost metric per link, instead of multiple link costs. The performance of Rahyab algorithm in terms of blocking rate is presented. In the proposed IA-RWA algorithm, the Quality of Transmission (QoT) of a lightpath is calculated using the DICONET Q-Tool. However, this tool, as well as any other QoT estimator, has only a certain degree of accuracy and therefore it introduces errors and inefficiencies in the lightpath selection process. We also show how the Rahyab multi-constraint algorithm can be extended to handle the inaccuracies of the Quality of Transmission (QoT) estimators. In particular, we consider an inaccuracy margin parameter per link that depends on the availability of particular monitors on that link. Inaccuracies may lead to accept lightpaths for which the QoT is actually too low, but estimated as satisfying. We show via simulations that in case of inaccuracies, we successfully block such connection attempts with poor actual QoT but a good estimated (when inaccuracies are not accounted for) QoT. In the second category of lightpath establishment techniques, the quality of transmission is estimated from probing information, rather than analytical models. In particular we leverage an end-to-end estimation framework called network kriging. The scheme relies on the exploitation of probing data from past lightpath establishments to estimate the Quality of Transmission (QoT) of new lightpaths before they are established. We use kriging to estimate end-to-end metrics such as Quality of Transmission (QoT) for a lightpath to be established, based on measured data from other, previously established or probed, lightpaths. Network kriging exploits the correlation between QoT metrics of lightpaths that share the same links in order to perform QoT estimation. In our lightpath establishment scheme, several attempts are performed to establish a lightpath. Our technique does not attempt to establish lightpaths with estimated poor QoT. We show that the technique is more appropriate to centralized control planes, although it is still applicable to distributed control planes. Lightpath establishment is only one of the building blocks of the core of the future. DICONET will integrate it into a more general framework to actually build a complete dynamic transparent optical including a control plane and failure localization modules. A brief overview of the future activities of the DICONET project, culminating with the building of a prototype, will be outlined.