Department  of  Compu.ng  and  Digital  System  Engineering

LASSU  Sustainability  on  ICT  Research  Laboratory  

Public  University,  founded  in  1934  

•   11  campi  (4  –  city  of  São  Paulo).  •   89  Uni8es.  

•   92.064  students  (undergrad,  grad  and  extension).  •   5.860  professors.  

•   16.837  administra8ve  staff.  •   249  undergratua8on  programs.    

• 239  gradua8on  Programs  

Source:  Anuário  Esta8s8co  2013  

University  of  São  Paulo

LASSU-PCS-EPUSP •  LASSU – Laboratory on Sustainability

•  Created in 2010 •  3 professors (Engineering School, Architecture & Urbanism, EACH) •  10 collaborators - Doctorate, Master and Undergrad students and

employees. •  Strong Partnership with CEDIR (Center For Discard and Reuse of E-

Waste)

•  Main fields of interest •  ITC Governance •  Network Management oriented to Sustainability Policies •  Energy Efficiency applied to:

•  SDN (Software Defined Network) •  Cloud Computing •  Data Centers

•  Electronic Waste •  Sustainability in Productive Chain •  Life Cycle Assessment

Main  Partnerships •  RNP  (Na8onal  Network  for  Research  and  Educa8on)  •  ANSP  (Academic  Network  of  São  Paulo  State)  •  I2Cat  –  Living  Laboratories  •  University  of  Illinois  –  Chicago  •  FIU  –  Florida  Interna8onal  University  

•  Ericsson  Research  Sweden,  Canada,  Finland  •  Center  for  InnovaJon  -­‐  Ericsson  Brazil  •  PETROBRAS  •  IBM  Research  –  T.J.  Watson  •  MIT  CISR  (Center  For  Informa8on  System  Research)  •  MIT  D-­‐Lab  •  MIT  L-­‐Lab  (Leadership  on  Sustainability)  

Sponsered  

Introduction •  Trade-off between performance and energy consumption à

challenge •  How to correlate quality of service and actions targeting

efficiency improvement? •  Is a broader sustainability view possible at network management

level? •  How to compare device efficiency and prioritize more efficient

device for multiple network paths? •  How to coordinate power-saving features in a heterogeneous

network?

SUStainability oriented Network Management System (SustNMS)

SWITC H  /  ROUTE RSWITC H  /  ROUTE RNE TWORK  MANAGEMENT  S Y S TEM

SUSTAINABILITY MONITOR

ENERGY EFFICIENCY EVALUATOR

POLICY MANAGEMENT FRAMEWORK

DEVICE UPDATER

QUALITY OF SERVICE MONITOR

MANAGEMENT PROTOCOL

AVAILABILITY EVALUATOR

TRAFFIC LOAD FORECAST

POL IC Y  MANAGEMENT  

TOOLS

POL IC Y  DE C IS ION  POINT

POL IC Y  E NFORC EMENT

PERFORMANCE EVALUATOR

MODE L  R E POS ITORY

AVAILABILITY MODEL

POWER MODEL

SWITC H  /  ROUTE R

SUBSYSTEMS

CARDS / PORTS

FORWARDINGENGINE / CPU

CHASSIS

MANAGEMENT SYSTEM CLIENT

MANAGEMENT PROTOCOL CLIENT

MIB

STATIC POWER METERS

TRAFFIC DATA

SUBSYSTEMS

CARDS / PORTS

FORWARDINGENGINE / CPU

AVAILABILITY DATA

POL IC Y  C ONS OLE

POLICY ENFORCEMENT POINT

POL IC IE S  R E POS ITORY

Results • S a v i n g s o f 4 3 % w h e n prioritizing energy efficiency

• Savings of 30% when no performance degradation is allowed

• Savings of 27% when a seven-nines reliability constraint is imposed

• Patent application

• 3 papers • 2 Master’s Thesis

Software Router (Linux-kernel based) MPLS-Linux support for TE

Power-saving mode emulation

SustNMS •  Monitoring through SNMP

•  Policy enforcement through CLI/SNMP •  Efficiency evaluation performed using power

profiles derived from the literature

Sponsored by Innovation Center – Ericsson Brazil

Introduction • ICT energy spending is growing

−  2% of worldwide emissions [1] • Networks represent a significant part of this amount

−  Voice and Data Networks ~23% in 2020 [1] • And they are usually overprovisioned • There are some research works aiming at saving energy in the network

−  Sleeping −  Link Rating

• How to bring business together? • How to coordinate the different green functionalities?

Current Research

• Energy efficiency features to save energy in the network à but there is no automated way to “bring” a sustainability high level policy to these features • There is no automated policy refinement method able to handle all the refinement requirements of sustainability-oriented policies

Expected Results • Comprise a method to support the selection of green functionalities using an utility function; • Use Table Lookup for policies translation; • Determine different information models to standardize the definition of policies and comply with the Policy Continuum; • Use parameterized policies to enable dynamicity, besides incorporating time conditions in the description of policies, and having a type of policy to change policies in case of change in scenario.

Sustainability Oriented System based on Dynamic Policies with Automated Policy Refinement (SOS)

Method

System

Busin

ess

Netw

ork

Devic

eInstance

Fase

Parameters,  attributes  and  constraints  identification  

CallForRefinement

The  BusinessPolicy  should  have  the  following  format:  user  (when  applicable),  target  (In  the  netwokr),  action  (GreenPlan,  save  energy),  

period  (night)  e  policy  validity  (6  months)

8.  Create  Policies(Rule  constructor)

Discover  managed  functionalities  in  each  device

A.  Fill  functionalities  details

SNMP  /  Web  /  Netconf

E.g.    -­‐  Save  energy  in  the  network  at  night                -­‐  Change  Energy  Source  In  the  netoork  in  the  morning

E.g.    -­‐Save  energy  in  the  network  from  9pm  to  5am  up  to  6  months

Business-­‐Level  Policy  Information  

Model

System-­‐Level  Policy  Information  Model

Information  model  Retrieval

1.Network  parameters,  attributes  and  constraints  

Identification  

Information  Retrieval

Existing  technologies  attibutes  and  parameters  

Useful  data  and  constrainst  to  configure  energy  saving  

In  the  network

Discover  resources  functionalities

B.  Verify  functionalities  availability

7.Network  Configuration  Evaluation  (Analytical  solver  /  Utility  function)

Network  Configuration  (List  of  actions  and  conditions)

9.  Verify  Policies(Goal  and  conflicts) Verify

If  needed,  Request  another  Network  configuration(Change  technologies  or  parameters)

4.  Workloads  generation(Random,  historical  

and  Forecast)

Workloads  todefine  possible  configurations

Device  parameters,  attributes  and  constraints  

Identification  

Create  Policy  for  the  devices

Device-­‐Level  Policy  Information  Model

Translate  to  commands

6.  Network  configurationgeneration  (functionalities  combination,  restrictions)

Possible  Network  Configurations  (List  of  actions  and  conditions)

Utility  function  parameters

Managed  functionalitiesdetails  (MFD)

Managed  functionalitiesdetails  (MFD)

Managed  functionalitiesavailable  and  combinations

Map  High-­‐Level  parameters  and  attibutes  to  System-­‐Level  

(Lookup  table)Policy  translation

Store  Policy

System-­‐Level  PoliciesSystem-­‐Level  PoliciesVerify  if  the  policy  exists

Business-­‐Level  PoliciesBusiness-­‐Level  Policies

Network-­‐Level  PoliciesNetwork-­‐Level  Policies

Store  Policies

Verify  if  such  a  policy  exists

CallForRefinement

Mapping  to  devicetechnologies Policy

Instance-­‐Level  PoliciesInstance-­‐Level  Policies

CallForRefinement

Available  Devices  Functionalities  Profile  (ADFP)Available  Devices  Functionalities  Profile  (ADFP)

Available  functionalitiesFunctionalities

5.Match  existing  functionalities  with  

the  available  functionalities

C.Check  functionalitiesconflicts

Network-­‐Level  Policy  Information  Model

Policy  console

InformationModelRetrieval

Store

Target,  time  condition

2.Check  network  topologyNetwork  Topology

Power  Profiles3.Check  devices’  power  profiles

ModelRepositoryModel

Repository

Discover  each  device  and  its  connections

Data input and initial translation

Network data gathering

Scenarios combination and selection using a Utility Function

Decision Tree generation and policies deployment

𝑼𝑭 = 𝒑𝒍   ∗ 𝟏 ∑ 𝑬𝒏𝒆𝒓𝒈𝒚𝑨𝒇𝒕𝒆𝒓𝑺𝒂𝒗𝒊𝒏𝒈𝒔𝑹𝒐𝒖𝒕𝒆𝒓  𝒌𝒏𝒌=𝟎∑ 𝑬𝒏𝒆𝒓𝒈𝒚𝑩𝒂𝒔𝒆𝒍𝒊𝒏𝒆𝑹𝒐𝒖𝒕𝒆𝒓  𝒌𝒏𝒌=𝟎

>  

Utility Function (our proposal)

Method

Sponsored by Innovation Center – Ericsson Brazil

Introduction •  Energy consumption demand

•  Datacenters •  Operators

•  New paradigm: Distributed Telecom Clouds

•  Objective: Expand SOS to consider automated policies in the context of Distributed Telecom Clouds and Smart Grids

Current Research

•  Sustainability-oriented policies refinement and energy efficiency functionalities management à SOS

•  Cloud platforms and virtualization

•  Smart Grids and renewable sources

Deliverables • Green service levels differentiation • Sustainability oriented policies specific to the Distributed Telecom Cloud Environment • Analysis of how Smart Grids can support • Expansion of SOS to the distributed telecom cloud environment • Specification and implementation of SustClouds

Sustainability Oriented Telecom Clouds with Automated Policy Refinement (SustClouds)

Continuation

Continuation

Continuation

Sustainable  practices

Rational  use  of  resources

Electrical    sourcessmart  grid

Compute Network Storage Cooling

Devices

Instances

SOS  scope

Continuation

Policy

refinement

Distributed  telco cloud  powered  by  renewable  energy  sources

Cloud orchestration extensions for  dynamicpower sources in  smart grids

Extensions  of  green  policy  refinement  for  integrating  new  types  of  resources  (physical  infrastructure)  and  associated  policies,  while  controlling  availability  trade-­‐offs

Sponsored by Innovation Center – Ericsson Brazil

LASSU  Main  Research  Areas  •  Energy  Efficiency  applied  to:  

•  SDN  (SoZware  Defined  Network)  •  Cloud  Compu8ng  •  Data  Centers  

•  Sustainability  Policies  for  ICT  System  •  Digital  technologies  as  Driver  for  Sustainability.  •  Smart  Grids.  •  ICT  Governance  •  Life  Cycle  Assessment  •  Sustainable  Produc8ve  Chain  

Thanks