The design and gradual deployment of all-optical\r\nbackbone network architectures will lead to the provisioning of\r\nenough bandwidth demands of future Internet of Things (IoT)�s\r\napplications and services. Current and future global traffic is increasingly\r\nbecoming heterogeneous as well as displaying structural\r\nself-similarity. Optical Burst Switching (OBS) is a practically\r\ncost-effective switching paradigm that can possibly address the\r\nmismatches between DWDM Terabit per second technology transmission\r\nspeeds and the required ultra-switching capabilities to\r\nserve the global traffic. Structurally, an OBS-switched backbone\r\nnetwork comprises buffered edge as well as bufferless core (interior)\r\nnodes. Because of the bufferless nature of the interior nodes,\r\nthe core of such networks tends to be quite prone to frequent contention\r\noccurrences that can easily cause significant overall network\r\nperformance degradation due to sporadic heavy burst loses.\r\nSuch drawbacks can be partly countered by appropriately dimensioning\r\navailable network resources, and reactively by way of deflecting\r\nexcess as well as contending traffic to available alternate\r\npaths. In doing so, diverted traffic must not trigger network performance\r\ndegradation in the deflection paths. Because minimizing\r\ncontentions is key to provisioning a consistent quality of service\r\n(QoS), we thus in this paper propose and analyze a scheme that\r\nseeks to intelligently deflect traffic in the core network such that\r\nQoS degradations caused by contention occurrences are negligible.\r\nSince contention minimization will lead to improved overall\r\nnetwork performance, we herein propose a regulated deflection\r\nrouting scheme (rDr) that utilizes neural network agents to reinforce\r\nthe deflection routing decisions at core nodes. The scheme is\r\nevaluated by way of simulations and results show that it can minimize\r\nburst loses as well as keep end-to-end delays for the various\r\napplications to within acceptable bounds. It also significantly reduces\r\nthe frequency of contention occurrences as well as burst deflections.

Username
Password