Presentation on theme: "Power-Aware Network Design"— Presentation transcript:
1Power-Aware Network Design «Power Awareness in Network Design and Routing»J. Chabarek et al.«Energy-Minimized Design for IP Over WDM Networks» G. Shen, R. S. Tucker
2Introduction The Internet is expanding tremendously Growth in the number of end users and connection speeds -> exponential increase in bandwidth demandIncrease in energy consumptionCost of transmission and switching one of the major barriersEnergy consumption may become a barrier1% - 2% of total electricity consumption in USA cut of 1% in the Internet energy consumption means about US$5 billion per yearIncrease in power densityThermal issues -> limitations of air coolingIncrease in operational costsIncrease greenhouse footprintSave the Earth!!!
3Power Aware Design Areas (I) Three main areas for power aware designSystem DesignDevelopment in CMOS technology -> improvements are slowing downMulti-Chassis Systems: separate physical components clustered forming a single logical routerAggregate power consumption increases -> heat spread over a large physical area -> existing cooling techniques usedAlternative Systems: optical switchesTerabits of bandwidth at much lower power dissipationProtocolsInvestigated in wireless networks -> Opportunities in wire-line networksBasic notion: put components to sleep if low traffic loadRouting protocols: routes calculated with power consumption constraints
4Power Aware Design Areas (II) Network DesignDeploy routers such that the aggregate power demand is minimizedSatisfying robustness and performanceTwo approachesMultiple router-level topologies satisfying capacity, robustness and power consumptionLimit power-hungry systems to a subset of routersSelection of chassis and line cards in routers is a main issue to reduce power consumptionIn IP over WDM networksIP routers use more than 90% of total powerLightpath bypass is used to reduce the number of IP router ports -> IP ports consume major energy in IP routers
5Router Power Consumption Router power consumption depends onType of router chassisType and number of line cards deployed in the chassisConfiguration and operating conditionsSize of packets100 bytes / 576 bytes / 1500 bytesSize of forwarding table1000 entries / entriesType of trafficUDPTCPEmployed protocols and techniquesOSPFNetflowUnicast Reverse Path Forwarding (uRPF)Access Control List (ACL)Active Queue Management - Random Early Detection (AQM – RED)
6Router Power Consumption Chassis and line card combinationsChassis: Cisco GSR / Cisco 7507
7Router Power Consumption Chassis and line card combinations (cont.)Base system is the most consuming7507 chassis + router processor -> 210 WattsGSR chassis + router processor + switching fabric -> 430 WattsIt is best to minimize the number of chassis and maximize the number of line cards per chassisCalculated power consumption of different cards
8Router Power Consumption Configuration and operating conditionsA 4-port Gigabit Ethernet line card and a OC-48 card in a GSR chassis is usedDeployed testbed:
9Router Power Consumption Configuration and operating conditions (cont.)Constant bit rate UDP traffic and different packet sizes1500 bytes / 576 bytes / 100 bytesPower consumption increases as packets get smaller!!!
10Router Power Consumption Configuration and operating conditions (cont.)Constant bit rate UDP traffic, medium packets and different featuresLarge forwarding table / ACL / uRPF / OSPFuRPF is the most consumingLarge forwarding table is less consuming!!
11Router Power Consumption Configuration and operating conditions (cont.)Self-similar TCP traffic, 75% offered load and different featuresNetflow / AQM - REDPower consumption similar to UDP with large-sized packets
12Router Power Consumption Configuration and operating conditions (cont.)Maximum variation in previous slides -> 20 WattsExtrapolating a fully loaded chassis -> WattsLess significant than chassis/line card configurationGeneral Model:PC -> power consumption of routerX is a vector defining chassis type, line cards, configuration and traffic profileCC -> power consumption of a chassis typeN -> number of line cardsTP -> scaling factor (traffic utilization)LCC -> cost of line card
13Power Consumption Optimization Main focus: allocation of line cards and chassis in nodes to minimize power consumptionMixed-Integer resource allocation problem with multicommodity flow constraintsInputsNetwork with OSPF link weightsTraffic matrixLine card and chassis optionsOutputsHow each node should be provisionedMultipath routingImplemented with General Algebraic Modeling System (GAMS)
14Power Consumption Optimization Networks are taken from the Rocketfuel projectInferred weights and link latenciesLink weights -> calculate approximate bandwidths of each linkTraffic matrixes generated with a gravity modelThree additional random graphs with 12 nodes and varying number of directed edges (Waxman method)
15Power Consumption Optimization Network design problem: deploy different chassis/line card configurations such that provisioning requirements are satisfied and power consumption in minimizedTraffic is scaled for each origin-destination pair -> linear scaling factorVaries provisioning requirementsTraffic flows might be altered to put cards/chassis to sleep in low utilizationFirst scenario includes only GSR chassis and OC-48 line cardOnly 10 line cards allowed per chassisScaling factor varies from 0.1 to 40
16Power Consumption Optimization Other experiments relaxing line cards per chassis, chassis type and card types (not in the paper)Minimum power consumption -> chassis accommodating large numbers of line cards and line cards capacities that closely match demand
17Power Consumption Optimization Power savingsCompared to a non-power-aware network design (shortest path)Using a specific chassis (GSR) and line cards (OC-48 or 0C-12)OC-12 line cards achieve smaller savings -> more ingress/egress node portsCost of additional connectivity is zero as long as the number of ports does not require additional line cards
18IP Over WDM Network IP layer: Optical layer Core IP router aggregates data traffic from low-end access routersIP router ports consume major energy (forwarding process) -> number of IP ports as measure of total power consumptionOptical layerOptical switches interconnected with physical fiber linksMay contain multiple fibersEach fiber needs a pair of multiplexer/demultiplexerEach wavelength require a pair of transponders -> full wavelength conversion is assumedEDFA amplifiers are deployed on fiber links
19IP Over WDM Network Two implementation approaches Lightpath non-bypass All data carried by lightpaths is processed and forwarded by IP routersAll lightpaths incident to a node must be terminatedLightpath bypassIP traffic whose destination is not the intermediate node -> directly bypasses the intermediate routerSaves IP router ports
20Energy Consumption Optimization for IP over WDM Network design problem: design an energy-minimized IP over WDM networkServing all the traffic demandsWith a limited maximal number of wavelengths in each fiberWith a limited maximal number of IP router ports at each nodeInputsPhysical topology -> N nodes and E linksTraffic demand matrixNumber of wavelength channels per fiber and capacity of each wavelengthMaximal number of IP router ports at each nodeEnergy consumption of router ports, transponders and EDFAs
21Energy Consumption Optimization for IP over WDM The optimization problem is solved using a Mixed-Integer Linear Programming (MILP) model includingEnergy consumption of IP routers, EDFAs and transpondersLayout of EDFAsPorts for aggregating data from low-end routersMILP model minimizes also the number of network components -> could be used for cost-minimized IP over WDM networkThe computational complexity is highO(N4) variables and O(N3) constraintsHeuristics are needed for fast solution
22Energy Consumption Optimization for IP over WDM - Heuristics Direct Bypass: directly establish virtual links (lightpaths) whose capacity is sufficient to accommodate all the traffic demands between each node pairRouting of lightpaths -> shortest path routingSimpleCould lead to low capacity utilizationMulti-hop bypass: traffic demands between different node pairs could share capacity on common lightpathsElongate lengths of some IP traffic flowsFewer lightpaths -> fewer IP router ports
23Energy Consumption Optimization for IP over WDM - Heuristics Multi-hop bypass heuristic:
24Energy Consumption Optimization for IP over WDM - Setup Five study casesLinear relaxation of the MILP model -> lower boundMILP optimal designNon-bypass -> upper boundDirect bypassMulti-hop bypassInputsTraffic demand between each pair node:Uniform distribution within a certain range centered at an identical average
25Energy Consumption Optimization for IP over WDM – Test Networks NSFNETUSNET
26Energy Consumption Optimization for IP over WDM – Total Power Consumption NSFNETLarger topology ->higher power consumption,heuristics closer to lower boundNon bypass -> upper boundLP relax. -> lower boundLinear relationship between total power consumption and total traffic demand intensityUSNET
27Energy Consumption Optimization for IP over WDM – Power Consumption Saving NSFNETLarger topology -> higher savings,longer lightpaths bypassing morenodes -> fewer IP portsMulti-hop bypass heuristic performs better than direct bypass ->Small traffic flows are aggregatedUSNET
28Energy Consumption Optimization for IP over WDM – Component Consumption NSFNET
29Energy Consumption Optimization for IP over WDM – Geographical Distribution NSFNETAll bypass design have a more uniform power distributionSolve problems associated with:Supplying large amounts of energyRemoving associated heat
30Energy Consumption Optimization for IP over WDM – Cost Analysis The model could be used for minimizing costChanging the optimization weights from energy to costMay NOT be valid if components with low energy consumption are the most expensive onesN6s8 network based on the MILP optimization model
31Conclusions Energy consumption may become a barrier for the Internet Operational costsGreenhouse footprintCooling issuesSupplying large amounts of energyPower aware design could solve itPower aware system designPower aware protocolsPower aware network designPower aware network design could achieve important savingsIn IP over WDM networks, lightpath bypass could save power consumption
32References[CHA08] J. Chabarek et al., «Power Awareness in Network Design and Routing», Proc. Of IEEE INFOCOM, 2008[SHE09] G. Shen, R. S. Tucker, «Energy-Minimized Design for IP Over WDM Networks», Journal of Optical Communication Networks, June 2009.