Periplus: A Resilient In-band SDN Control Plane via Embedded Forwarding Graphs
Networking and Internet Architecture
Summary
The authors developed Periplus, a system to help manage networks where devices have limited resources and are spread out over large areas. It solves problems like starting up the network without flooding it with messages, routing data efficiently, and quickly fixing problems when connections fail. Periplus works by letting switches handle backup routes on their own without needing to ask the controller every time something goes wrong. The authors tested Periplus on simulated networks and found it to be fast, scalable, and efficient in using switch memory.
Authors
E. M. Castro Barbero, P. de las Heras Quirós, F. J. Simó Reigadas
Abstract
Many resource-constrained, wide-area telecommunications deployments could benefit from an in-band SDN control plane, but several challenges stand in the way. This paper presents Periplus, an in-band SDN control plane designed to address four challenges that this approach presents in such contexts: automatic bootstrapping, source-based routing, fast failure recovery, and multi-controller coordination. The first three are developed in detail, while multi-controller coordination is addressed in a separate paper. For bootstrap, Periplus avoids network-wide flooding: when a new switch joins, the controller installs flow rules in only two switches. For routing and failure recovery, Periplus encodes a primary path and per-hop alternatives in a graph structure encapsulated between L2 and L3 headers; switches forward along the primary path and, upon detecting a failure, fall over locally to the encoded alternative without controller involvement. We evaluate a Ryu-based implementation of Periplus in Mininet across multiple topologies. Periplus runs on stock Open vSwitch (OVS), relying only on its built-in Nicira extensions for Network Service Header (NSH) encapsulation. The evaluation shows sub-50 ms failure recovery, scalable bootstrap across topologies of varying size and diameter, and a per-switch flow-table footprint that is independent of network size and grows only at switches where the controller encodes multiple alternatives.