Routing, modulation, spectrum and launch power assignment to maximize the traffic throughput of a nonlinear optical mesh network

Abstract

We investigate the optimization of routing, modulation format adaptation, spectral and launch power assignment as a means of improving the utilization of limited network resources and increasing the network throughput. We consider a transparent optical network operating in the nonlinear transmission regime and using the latest software adapted coherent optical techniques. We separate the problem into one of routing, modulation adaption and channel assignment, followed by channel spectral assignment, and launch power allocation. It is shown, for three test networks, that the launch power allocation and channel spectral assignment can improve the transmission SNR margin over the fixed modulation, fixed power, fully loaded link worst case by approximately 3-4 dB. This increase in SNR margin can be utilized through modulation format adaption to increase the overall network throughput. This paper highlights that increased gains in network throughput can be achieved in nonlinear impaired networks when individual transmitter spectral assignment and launch power are optimized to minimize the nonlinear interference.

Estimating Network Throughput with an Adaptive Routing and Wavelength Assignment Algorithm

“…Optimal network capacity was found using the MILP described in [3]. This models a transparent, fixed-grid network and assigns modulation formats according to signal-to-noise ratio, SNR, estimates for the worst-case, central WDM channel.…”

Joint Assignment of Power, Routing, and Spectrum in Static Flexible-Grid Networks

“…We consider three topologies for the numerical comparison, a simple 6-node network in Figure 3 and the German and NSF networks [18]. The fiber power attenuation α = 0.22 dB•km −1 , the nonlinear parameter γ = 1.3 W −1 • km −1 , the group velocity dispersion β 2 = −21.3 ps 2 • km −1 , the spontaneous emission factor n sp = 1.58 (i.e., the noise figure is 5 dB), the light frequency ν = 193.55 THz and the length of each fiber span L is 100 km.…”

“…The highbit-rate requests correspond to Nyquist-WDM superchannels formed by multiple subcarriers without guardbands between each other to enable a higher spectral efficiency [20]. The uniform random traffic profile is a widely used assumption [17], [18], [32], [33] to facilitate the numerical evaluation. For each topology, 300 random traffic matrices are used in the simulation.…”

“…This is because the average link length is shorter and, thus, the connections can choose higher-order modulation formats in the German network. As the modulation orders are higher, the SNR thresholds grow exponentially in a linear scale [18] and the margins between the actual SNRs in the benchmark and the thresholds become bigger. Therefore, there is a larger room to reduce connection PSDs for the proposed heuristic to minimize the overall NLI without affecting the spectrum utilization.…”

“…Power optimizations in optical networks with precalculated routing and spectrum assignment (RSA) schemes have been proposed [13]- [15], which yield only suboptimal resource allocation due to the separate optimization of power and other resources. Joint individual power and RSA optimization in nonlinear WDM networks [16]- [18] demonstrating improved network throughput have also been studied. However, these algorithms are designed specifically for WDM networks and cannot be applied to flexible-grid networks where connection bandwidths are not uniform.…”