← Research

SOS: Sustainability Oriented Policies based on Dynamic Policies with Automated Policy Refinement

Bruno Rodrigues' contribution to a University of São Paulo and Ericsson Research project: orchestrating a network's energy-saving features from business-level sustainability policies, tested on GreenSDN, the energy-aware SDN testbed he built.

StatusCompleted
Period2014 to 2015
PartnersUniversity of São Paulo (USP), Laboratory of Sustainability in ICT (LASSU), Ericsson Research
Contribution byBruno Rodrigues

Network equipment had gained many ways to save energy: links that slow down when traffic is light, interfaces that sleep between bursts, routing that packs traffic onto fewer devices so others can be switched off. Each one was designed on its own, for one layer of the network. Switched on together, they can add up to larger savings, or work against each other, and a feature that suits a quiet network causes congestion and packet loss in a busy one. Operators had no systematic way to decide which ones to run, when, and in which combination.

SOS addressed that problem. It was a joint project of the Laboratory of Sustainability in ICT (LASSU) at the University of São Paulo, led by Prof. Tereza Carvalho, and Ericsson Research, with Catalin Meirosu. The method at its core was the doctoral work of Ana Carolina Riekstin (PhD thesis, USP, 2016). Bruno Rodrigues joined toward the end of the project, as a Master’s student, and contributed the testbed on which SOS was run. This page describes that contribution, not the wider project.

The method: from business policy to network action

SOS starts from policy-based network management: an operator writes high-level goals, for example to save energy without breaking a performance target, and a refinement process translates them down to concrete settings on the devices. The project surveyed existing policy refinement methods and the requirements that sustainability goals add to them (IEEE Communications Surveys & Tutorials, 2016). The orchestration method then combines the available energy efficiency capabilities while avoiding the combinations that conflict, using a utility function to weigh energy savings against performance and a decision tree that picks the best option for the current bandwidth utilisation, topology, and power profile of the devices.

GreenSDN: a testbed for energy-saving networks

The emulated topology, inspired by the Brazilian research network (RNP), with traffic sources and sinks at its edges. From the GreenSDN Master’s thesis (2016).
The emulated topology, inspired by the Brazilian research network (RNP), with traffic sources and sinks at its edges. From the GreenSDN Master’s thesis (2016).
A method that orchestrates energy-saving features needs a network where those features actually exist, and experimental environments designed for that were lacking. GreenSDN filled the gap. Built on the Mininet emulator and the POX OpenFlow controller, it emulates the effects of energy efficiency capabilities at three scopes of the network: the interface, the device, and the network as a whole. Experiments ran on an emulated topology inspired by the Brazilian research network (RNP), comparing the savings obtained by activating each capability. GreenSDN was published at IFIP/IEEE IM 2015, where SOS was also demonstrated running on it, and became the subject of Bruno’s Master’s thesis (GreenSDN: Energy Efficiency in Software Defined Networks, USP, 2016, PDF).

GreenSDN’s architecture and workflow: SOS and a GUI on top, a management layer that refines policies into decisions, power emulation of the energy saving capabilities (ALR, SC) and QoS monitoring in the POX control plane, and Open vSwitch nodes in Mininet as the data plane. From the GreenSDN Master’s thesis (2016).
GreenSDN’s architecture and workflow: SOS and a GUI on top, a management layer that refines policies into decisions, power emulation of the energy saving capabilities (ALR, SC) and QoS monitoring in the POX control plane, and Open vSwitch nodes in Mininet as the data plane. From the GreenSDN Master’s thesis (2016).

The experiments confirmed the project’s premise: orchestrating several capabilities together yields additional savings over any single one, while keeping their operation free of conflicts.

Publications

What the group builds on it

SOS was Bruno’s first encounter with two ideas that kept returning. One is policy-based management: stating what an operator wants once and letting the system choose how, which reappeared in BC4CC as policy-driven blockchain selection. The other is the value of a testbed that makes an idea measurable before it meets a production network, an approach the group still follows with its WiFi sensing mesh and its plant deployments. The follow-up project, E2C, carried SOS into cloud data centres.