Open Networking & Edge Summit 2020

eBay's network and central operations teams need the ability to observe its massive network infrastructure to meet eBay's availability goals.

While tools are available to detect hard device failures, the majority of issues in this infrastructure arise from "gray failures". These result in either application latency or failures that are detrimental to smooth operations.
Given the complexity of our network, these "gray failures" are a non-trivial problem to detect.

NetVision is an agent-based, end-to-end observability platform for context-based data plane monitoring of eBay's network. It is designed to identify partial packet losses along with hard device failures in the network and is particularly useful in identifying the location of the issue in the complex network topology.
Using a combination of end-to-end UDP probing, master/agent architecture and intelligent algorithms, NetVision has been effective in identifying network issues quickly thereby reducing the time to detect issues.