Network Flow Data in the Cloud#

Steve Dodd, Slack — Core

• Why: An argument that classical network-engineering techniques (graph theory, the discrete math underlying traditional routing/traffic-engineering work) are directly applicable to modern cloud service-oriented architecture — and a concrete method for per-service traffic attribution without paying for vendor flow-logging products.
• What: A framework for building per-service network traffic attribution in cloud infrastructure using network-graph analysis techniques borrowed from traditional network engineering (the kind of work "manual OSPF metric tuning" represents), without relying on expensive vendor logging solutions. (from abstract only)
• How: Applies graph-theoretic analysis of network topology/flow to optimize data flow, routing, and resilience visibility at the service level — essentially traffic engineering applied to service-mesh-era cloud infrastructure. (from abstract only)
• Where applicable: General for large-scale cloud service infrastructure wanting traffic/dependency visibility; the technique is explicitly framed as reviving under-used traditional networking analysis for cloud-native environments.