期刊名称：International Journal of Computer Science and Network Security
印刷版ISSN：1738-7906
出版年度：2020
卷号：20
期号：12
页码：193-201
DOI：10.22937/IJCSNS.2020.20.12.21
出版社：International Journal of Computer Science and Network Security
摘要：This paper proposes a novel wavelength and bandwidth resource allocation scheme for NG-EPON, where the resource allocation is completely adaptive based on the traffic reported by each optical network unit (ONU) in the network. NG-EPON supports up to 100 Gb/s data rate with 4 wavelengths with an individual capacity of 25 Gb/s. The proposed dynamic wavelength and bandwidth allocation (DWBA) scheme is based on online gated service to assign adaptive number of wavelengths and dynamic bandwidth size on each selected wavelength. The resource assignment is done such that it not only fulfills the bandwidth needs of each ONU in the network but also utilizes all the wavelength resources in equal and effective manner. First the bandwidth is allocated on the wavelength with the least utilization and new wavelength is allocated only if its bandwidth utilization is more than the bandwidth overhead, i.e. guard time. Further, this effectiveness is evaluated with the help of a parameter named alpha, and its optimal value is decided based on machine learning. The proposed DWBA algorithm is evaluated in terms of average packet delay, grant utilization, and number of wavelengths used in comparison with modified-IPACT, First-Fit DWBA (FF-DBA), and water-filling algorithm based DWBA (WF-DBA) algorithms. The proposed DWBA performance is validated with the help of simulation results.
关键词：NG-EPON; dynamic wavelength and bandwidth allocation; average packet delay; machine learning.
