US 11,695,682 B2
Optimizing Border Gateway Protocol (BGP) traffic using reinforcement learning
Cengiz Alaettinoglu, Sherman Oaks, CA (US); Shelley A. Bhalla, Morrisville, NC (US); Emil Janulewicz, Ottawa (CA); Thomas Triplet, Manotick (CA); and David Côté, Gatineau (CA)
Assigned to Ciena Corporation, Hanover, MD (US)
Filed by Ciena Corporation, Hanover, MD (US)
Filed on Sep. 30, 2019, as Appl. No. 16/587,404.
Prior Publication US 2021/0099378 A1, Apr. 1, 2021
Int. Cl. H04L 45/02 (2022.01); G06F 30/20 (2020.01)
CPC H04L 45/08 (2013.01) [G06F 30/20 (2020.01); H04L 45/04 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A system configured to optimize Border Gateway Protocol (BGP) traffic in a telecommunications network, the system comprising:
a network interface configured for communication with the telecommunications network;
a processing device interconnected with the network interface; and
a memory device configured to store instructions that, when executed, enable the processing device to
perform an action in the telecommunications network when one or more inter-Autonomous System (AS) links are in a current state, wherein the action in the telecommunications network is configured to have an effect on BGP traffic flow on the one or more inter-AS links,
define one or more metrics, describing a state of the one or more inter-AS links, based on the effect of the action on the BGP traffic flow,
determine an updated current state of the one or more inter-AS links based on the defined one or more metrics, and
utilize the defined one or more metrics to perform a further action to achieve one or more rewards associated with the one or more inter-AS links, wherein an offline simulation is utilized to determine which actions are effective for the one or more rewards based on the current state, the actions and the one or more rewards are based on reinforcement learning where the current state is mapped to some action to maximize the one or more rewards over time, wherein the reinforcement learning seeks to learn what to do given a problem, in the BGP traffic flow, for the current state being mapped to the some action.