FlowState
Replaying simulated run Source

Traffic that reroutes itself.

FlowState is an SDN controller that classifies every flow, forecasts which links are about to congest, and sends new traffic another way. Below is one simulated run: every train is a real flow on the route its controller chose.

Campus network

T00 / 23
Voice Video Web Bulk

–
lower average latency
–
lower peak link load
–
congested flows, static to adaptive
–
traffic classifier accuracy (held-out)
From the Python experiment: 24 controller intervals, identical traffic for both controllers.

Interval board

Busiest link in each interval. Select one to jump there.

    Average flow latency

    Milliseconds, Python experiment

    StaticAdaptive

    Peak link utilization

    Percent; over 100 means over capacity

    StaticAdaptive

    How the controller decides

    Five stops, repeated for every new flow

    1. Read the network

      Poll flow statistics and per-port byte counters from every switch.

    2. Classify the flow

      A Random Forest labels it voice, video, web or bulk, which sets its priority.

    3. Forecast load

      A second model predicts each link's utilization for the next interval.

    4. Weigh every link

      Busy and soon-to-be-busy links get expensive, more so for voice and video.

    5. Install the path

      Dijkstra picks the cheapest path and the controller pushes OpenFlow rules hop by hop.

    cost = base delay + priority factor × (35 × current load³ + 12 × forecast load³)