Error Awareness in a Multi-Layer Transport Network OS

Software-defined control with a network operating system combined with flexible-bitrate optical transponders allows near zero-margin operation of optical networks to "squeeze" as much capacity out of a link as possible. Since their inception, optical transport networks have been designed with the assumption that operational optical links must essentially be error free after applying forward error correction. However, operating close to zero system margin increases the likelihood that error bursts may occur on a link. In this work, we explore the possibility of utilizing optical links with a low level of burst errors to carry error-tolerant services. This requires monitoring the amount of link errors so that new error-aware service level agreements can be met. To determine acceptable error levels for a service, we introduce and demonstrate the measurement of service-specific performance maps, such as the flow completion-time map, based on physical-layer parameters such as the long-term average biterror ratio. This allows us to demonstrate error-aware intents in a multi-layer transport network operating system with service differentiation so that error-intolerant traffic can be migrated away from a link with errors while keeping the more errortolerant traffic on the link.We specifically study the application of data-center interconnection as a service with potential tolerance to links with errors and measure the flow completion times and ping loss ratios for flows on such a link. On our testbed network, we show that the data-center interconnection throughput can be essentially doubled over a range of 7 orders of magnitude in biterror ratio compared to the present mode of network operation.