Magellan: Generating Multi-Table Datapath from Datapath Oblivious Algorithmic SDN Policies

Abstract Despite the emergence of multi-table pipelining as a key feature of next-generation SDN data-path models, there is no existing work that addresses the substantial programming challenge of utilizing multi-tables automatically. In this paper, we present Magellan, the first system that addresses the aforementioned challenge. Introducing two novel, substantial algorithms, map-explore and table-design, Magellan achieves automatic derivation and population of multitable pipelines from a datapath-oblivious, high-level SDN program written in a general-purpose language. Comparing the flow tables generated by Magellan with those produced from standard SDN controllers including OpenDaylight and Floodlight, we show that Magellan uses between 46-68x fewer rules.
Authors
  • Andreas Voellmy (Yale)
  • Shenshen Chen
  • Xin Wang
  • Richard Yang (Yale)
Date Aug-2016
Venue Annual Conference of the Special Interest Group on Data Communication 2016