Optimization, Design, and Analysis of Flexible-Grid Optical Networks with Physical-Layer Constraints
Doktorsavhandling, 2018
The theme of this thesis is the optimization, design, and analysis of flexible-grid optical networks that are constrained by physical-layer impairments (PLIs). We consider three flexible-grid network scenarios. The networks in the first class are static nonlinear transparent backbone networks where physical-layer resources are allocated to each traffic demand. The networks in the second class are traffic-variable nonlinear translucent backbone networks where regenerator sites are necessary to recover optical signals from the accumulated noise in long-distance transmission. The third class is data-center networks based on optical spatial division multiplexing. Within each class, our focus is primarily on an efficient and balanced allocation of network resources. Both optimization formulations and heuristic algorithms are proposed for each class. The contributions of this thesis can thus be categorized into three topics, as outlined below.
First, we consider the optimization of network resources in the presence of PLI. The PLI between optical connections is characterized by the Gaussian noise (GN) model and incorporated into resource allocation algorithms. As an example, for a link-level optical communication system, the spectrum usage can be reduced by roughly up to 22% by accurately modelling the PLIs and assigning proper modulation formats and spectrum to optical connections. For resource allocation in the network level, the power spectral density of each optical connection is optimized in addition to the previously mentioned resources.
As a second topic, the design of flexible-grid optical networks is studied. Specifically, we consider the regenerator location problem in traffic-variable translucent backbone networks. Due to the constantly changing traffic, the PLIs suffered by optical connections are also stochastic and, thus, have to be handled from a probabilistic perspective. A statistical network assessment process is used to characterize the noise distributions suffered by optical connections on each link, based on which a heuristic algorithm is proposed to select a set of regenerator sites with the minimum blocking probability.
Finally, we study the trade-off between the blocking probability and total throughput in the modular data center networks (DCNs) based on different optical spatial division multiplexing switching schemes. This performance trade-off is caused by the coexistence of traffic demands with extremely different data rates and number of requests in DCNs. A heuristic resource allocation algorithm is proposed to enable flexible tuning of the objective function and achieve a balanced network performance.
First, we consider the optimization of network resources in the presence of PLI. The PLI between optical connections is characterized by the Gaussian noise (GN) model and incorporated into resource allocation algorithms. As an example, for a link-level optical communication system, the spectrum usage can be reduced by roughly up to 22% by accurately modelling the PLIs and assigning proper modulation formats and spectrum to optical connections. For resource allocation in the network level, the power spectral density of each optical connection is optimized in addition to the previously mentioned resources.
As a second topic, the design of flexible-grid optical networks is studied. Specifically, we consider the regenerator location problem in traffic-variable translucent backbone networks. Due to the constantly changing traffic, the PLIs suffered by optical connections are also stochastic and, thus, have to be handled from a probabilistic perspective. A statistical network assessment process is used to characterize the noise distributions suffered by optical connections on each link, based on which a heuristic algorithm is proposed to select a set of regenerator sites with the minimum blocking probability.
Finally, we study the trade-off between the blocking probability and total throughput in the modular data center networks (DCNs) based on different optical spatial division multiplexing switching schemes. This performance trade-off is caused by the coexistence of traffic demands with extremely different data rates and number of requests in DCNs. A heuristic resource allocation algorithm is proposed to enable flexible tuning of the objective function and achieve a balanced network performance.
GN model
optimization
resource allocation
Flexible-grid optical network
heuristic algorithm
DCNs
regenerator location
EB lecture hall
Opponent: Associate Professor Vittorio Curri, Department of Electronics and Telecommunications, Politecnico di Torino
Författare
Li Yan
Chalmers, Elektroteknik, Kommunikation, Antenner och Optiska Nätverk
Link-Level Resource Allocation for Flexible-Grid Nonlinear Fiber-Optic Communication Systems
IEEE Photonics Technology Letters,;Vol. 27(2015)p. 1250-1253
Artikel i vetenskaplig tidskrift
Resource Allocation for Flexible-Grid Optical Networks With Nonlinear Channel Model
Journal of Optical Communications and Networking,;Vol. 7(2015)p. B101-B108
Artikel i vetenskaplig tidskrift
Joint Assignment of Power, Routing, and Spectrum in Static Flexible-Grid Networks
Journal of Lightwave Technology,;Vol. 35(2017)p. 1766-1774
Artikel i vetenskaplig tidskrift
Robust regenerator allocation in nonlinear elastic optical networks with time-varying data rates
Network performance trade-off in modular data centers with optical spatial division multiplexing
Optical communication systems offer tremendous potential capacity in long distances and, thus, are the foundation of today’s information society that connecting everyone in the world. Every time a web page is accessed, a phone call is made, or an email is sent, optical communication systems are probably involved at some point on the way from and to the user.
Even though optical fibers provide very high-speed communications, the resources of the optical network are still relatively limited, especially considering that the large data volumes generated by our society is rapidly increasing. On the other hand, driven by the growing traffic, a huge amount of fiber links and devices in the optical network need to be properly controlled. For example, network designers need to determine how to route traffic through the network, how to ensure the resource contention issues are minimal, and how to improve the reliability of services. It is prohibitively difficult to manually plan such a large and complex network. Therefore, efficient resource allocation algorithms become essential in handling traffic demands and producing cost-effective network configurations.
To achieve efficient utilization of network resources, multiple traffic demands usually share a common fiber link along their routes in the network. However, the coexistence of traffic demands can disturb the data transmission by interfering with each other. If the interference is beyond some acceptable levels, the information sent by the transmitter will not be correctly understood by the receiver, therefore leading to unsatisfactory quality of service. To avoid this, the optical network need to be appropriately configured such that the interference between traffic demands is minimized.
In this thesis, we are concerned with designing efficient resource allocation algorithms for optical networks affected by interference between traffic demands. We first develop resource allocation algorithms that mitigate the interference in link-level optical communication systems. These algorithms are then generalized to the network level to handle more traffic demands and resources. Furthermore, we propose resource allocation algorithms for other optical networking scenarios, such as the performance analysis of optical data center networks and the placement of regenerators in backbone optical networks for signal quality recovery. The interested reader is referred to the page i of the thesis for a technical abstract summarizing the contributions.
Even though optical fibers provide very high-speed communications, the resources of the optical network are still relatively limited, especially considering that the large data volumes generated by our society is rapidly increasing. On the other hand, driven by the growing traffic, a huge amount of fiber links and devices in the optical network need to be properly controlled. For example, network designers need to determine how to route traffic through the network, how to ensure the resource contention issues are minimal, and how to improve the reliability of services. It is prohibitively difficult to manually plan such a large and complex network. Therefore, efficient resource allocation algorithms become essential in handling traffic demands and producing cost-effective network configurations.
To achieve efficient utilization of network resources, multiple traffic demands usually share a common fiber link along their routes in the network. However, the coexistence of traffic demands can disturb the data transmission by interfering with each other. If the interference is beyond some acceptable levels, the information sent by the transmitter will not be correctly understood by the receiver, therefore leading to unsatisfactory quality of service. To avoid this, the optical network need to be appropriately configured such that the interference between traffic demands is minimized.
In this thesis, we are concerned with designing efficient resource allocation algorithms for optical networks affected by interference between traffic demands. We first develop resource allocation algorithms that mitigate the interference in link-level optical communication systems. These algorithms are then generalized to the network level to handle more traffic demands and resources. Furthermore, we propose resource allocation algorithms for other optical networking scenarios, such as the performance analysis of optical data center networks and the placement of regenerators in backbone optical networks for signal quality recovery. The interested reader is referred to the page i of the thesis for a technical abstract summarizing the contributions.
Styrkeområden
Informations- och kommunikationsteknik
Ämneskategorier
Telekommunikation
Kommunikationssystem
Infrastruktur
C3SE (Chalmers Centre for Computational Science and Engineering)
ISBN
978-91-7597-763-8
Doktorsavhandlingar vid Chalmers tekniska högskola. Ny serie: 4444
Utgivare
Chalmers
EB lecture hall
Opponent: Associate Professor Vittorio Curri, Department of Electronics and Telecommunications, Politecnico di Torino