Tomography Based Learning for Load Distribution Through Opaque Networks

Applications such as virtual reality and online gaming require low delays for acceptable user experience. A key task over-the-top (OTT) service providers who provide these applications is sending traffic through the networks to minimize delays. OTT typically generated from multiple data centers which are multi-homed several network ingresses. However, information about path characteristics of underlying ingresses destinations not explicitly available services. These can only be inferred external probing. In this paper, we combine tomography with machine learning We consider problem in a general setting where sources choose set their enter black box network. The viewed reinforcement strict linear constraints on continuous action space. Key technical challenges solving include high dimensionality handling that intrinsic networks. Evaluation results show our methods achieve up 60% delay reductions comparison standard heuristics. Moreover, develop used centralized manner or distributed by independent agents.