On Flow Concurrency in the Internet and its Implications for Capacity Sharing

Flow concurrency is the measure of the number of active flows at a given point in time at a given point in the net-work, and can be used as a complement to traffic volume to understand the dynamics of a measured network. It is of particular interest given the spread of devices through the network which keep per-flow state. In this work, we first present a simple methodology for measuring flow concur-rency using network flow data, then apply this methodol-ogy to a long-term data archive captured at the border of a national-scale research network to measure flow concur-rency in selected example network configurations (e.g., at a content consumer network, on a content provider network, at an interconnect point). Flow concurrency is interesting in the context of capac-ity sharing efforts in two ways. First and most obviously, devices which verify and enforce policy compliance for ca-pacity sharing must keep per-flow state on either end of a flow; flow state requirements therefore dictate where such devices may be placed in the network, and the trust proper-ties of the algorithms they run. Second, as the deployment of flow-state-keeping devices in the network increases, flow state itself becomes a “congestible ” resource just as queue space is: future work in capacity sharing may consider ad-dressing this