With the rapid adoption of the Internet of Things and mobile devices, the global Internet traffic is estimated to increase by nearly three times from 2017 to 2022. Hence, the demand for higher bit rates is necessary to support massive data transfers over optical networks. Ideally, a network that is able to handle many long distance requests at high transmission speeds is preferable. However, given limited spectrum resources, a trade-off is inevitable when selecting the correct routing parameters. On the one hand, we can minimize the usage of resources using spectrally efficient modulation formats. As a result the effective transmission distance of the signal will be reduced. Similarly, if we are to use a low-level modulation format, we can increase the effective range of transmission, but the amount of the used spectrum will also increase. It is in these scenarios where the importance of regenerators become apparent. When the link distance between two nodes is too large, a regenerator is required to regenerate the signal strength. However, regenerators are expensive to install. Additionally, given that the regenerator number needs to be set during the design of the network architecture, efficient calculation of their allocation is required to reduce each node's limitations, thus optimizing the network CAPEX and OPEX. In this paper, we demonstrated the feasibility of using deep tensor neural network to optimize regenerator placement in optical networks. Our approach brings significant improvements over the results achieved by the currently deploying techniques and can be used in the real-life networks.
Paper
Full text
Regenerator Placement in Survivable Optical Networks Using Deep Tensor Neural Network
Semantic Scholar · Engineering · 2020
Abstract
With the rapid adoption of the Internet of Things and mobile devices, the global Internet traffic is estimated to increase by nearly three times from 2017 to 2022. Hence, the demand for higher bit rates is necessary to support massive data transfers over optical networks. Ideally, a network that is able to handle many long distance requests at high transmission speeds is preferable. However, given limited spectrum resources, a trade-off is inevitable when selecting the correct routing parameters. On the one hand, we can minimize the usage of resources using spectrally efficient modulation formats. As a result the effective transmission distance of the signal will be reduced. Similarly, if we are to use a low-level modulation format, we can increase the effective range of transmission, but the amount of the used spectrum will also increase. It is in these scenarios where the importance of regenerators become apparent. When the link distance between two nodes is too large, a regenerator is required to regenerate the signal strength. However, regenerators are expensive to install. Additionally, given that the regenerator number needs to be set during the design of the network architecture, efficient calculation of their allocation is required to reduce each node's limitations, thus optimizing the network CAPEX and OPEX. In this paper, we demonstrated the feasibility of using deep tensor neural network to optimize regenerator placement in optical networks. Our approach brings significant improvements over the results achieved by the currently deploying techniques and can be used in the real-life networks.