Fujitsu Laboratories Ltd. and Fujitsu Laboratories of America, Inc. have developed a new technology to estimate the transmission performance as well as improve the throughput of optical networks.
Until now, it was difficult to determine the actual transmission performance of the optical fibers and communication devices. As a result, the networks are required to be operated under more restrictive conditions leading to lower network throughput than what the network is able to support.
However, the new technology from Fujitsu learn the actual characteristics of the optical network and observes the bit error ratio (BER) of the communication paths already in operation, thereby estimating the supported transmission speed of the newly added paths.
The technology was experimentally evaluated on a testbed simulating an optical network with the largest transmission distance of about 1,000 km. The results confirmed that the estimated erred by less than 15% and the throughput of the network could be improved as a whole by around 20%.
Overview of old technology compared to Fujitsu’s new Technology
The present situation requires meeting the increasing communication demands accompanied by expansion of the cloud-based services and the commercialization of the 5G mobile communication systems. Consecutively, there is going to be an unprecedented demand for technologies that can increase the throughput of optical networks.
Fujitsu’s newly developed technology estimates transmission performance by studying network characteristics using BER, observed by the equipment making up an operating optical network. In this technology, the BER based on the physical characteristics of the optical network is compared to the one obtained from optical receivers in operation. Then the difference is minimized by using machine learning on the mathematical model.
The new technology analyses each response of BER like loss in optical fibers or amount of noise in the amplifiers. This helps the system automatically calculate an appropriate amount of parameter value adjustment merely from the BER data in response to the physical parameters, which in turn renders efficient machine learning.
Filed Under: News