Understanding what is currently going on in a network is very difficult. Even simple questions such as “How is my network forwarding Google traffic?” often require network operators to manually bridge a large semantic gap from the underlying data to the actual answer. Bridging this gap is complex as the operator has to infer high-level insights from a wealth of low-level forwarding rules and traffic statistics distributed across many routers. This process often takes hours to complete (if at all). Yet, fast reasoning is essential when it comes to debugging networks, especially as today's networks carry more and more critical services. Wouldn't it be great, if you could simply ask your network such questions and it would automatically give you an answer to it? We envision virtual assistants for network operators that, akin to the ones typically found in smartphones, support operators in their daily work. A network virtual assistant takes on the time-consuming, tedious tasks such as processing, refining, and summarizing the low-level data. The operator automatically gets for all her questions summarized answers including all relevant insights, allowing her to focus on drawing the right conclusions. For example, the assistant would answer the above question with: “Google traffic follows the shortest-path to one of the two egresses: New York, Seattle”. In this talk, we show our first steps towards this vision and present a system that produces succinct summaries in natural language of the network-wide forwarding state guided by queries of the network operator. We show that the network can answer you and are curious to hear your questions!
Networks · Services · People | Code of Conduct | Privacy Policy
