Review of BS and AP management schemes
Transcript
Review of BS and AP management schemes
Review of BS and AP management schemes Łukasz Budzisz, Fatemeh Ganji, Adam Wolisz TU Berlin, Germany Gianluca Rizzo, IMDEA, Spain (currently with Univ. of App. Sciences of Western Switzerland) Marco Ajmone Marsan, Michela Meo, Yi Zhang, Politecnico di Torino, Italy Alberto Conte, Ivaylo Haratcherev, Alcatel-Lucent Bell Labs, France George Koutitas, Leandros Tassiulas, University of Thessaly, Greece Mario Pickavet Bart Lannoo, Sofie Lambert, Ghent Univ. - iMinds, Belgium Energy efficient wireless access networks n Design approach for wireless access networks: q n In reality, however: q q q n n Obtain maximum performance at full load most of time, the wireless access networks are under low or medium traffic load load profile exhibits large variation (close to the users) BS consumption is little load proportional and consume about the same at any load Savings between 20-40% can be achieved BS sleep modes are needed Sleep modes require BS management algorithms Michela Meo – Politecnico di Torino BS/AP management schemes - taxonomy Proposed framework points out: n the most important design aspects n shortcomings and advantages n energy-saving potential Michela Meo – Politecnico di Torino Existing BS management schemes n Flat Network: with single type of BSs, macrocells/ microcells only, one operator q q n Multi-Tier Network: with multiple types of BSs, also multiple technologies q q n Non-overlapping architecture Overlapping architecture Wi-Fi offloading Femtocells Mobile operator co-operation Michela Meo – Politecnico di Torino Flat network with non-overlapping architecture n n If the no. of BS always on is large, a slight increase of the RF output power is enough (if any) If the no. of BS always on is small, RF output power increase are needed, as well as titling adjustments Michela Meo – Politecnico di Torino Flat network with non-overlapping architecture Two control schemes: n Centralized approach à higher complexity of the decision, higher guarantee of reliable QoS n Distributed approach à BSs or groups of BSs takes decisions independently Michela Meo – Politecnico di Torino Flat network with overlapping architecture n n Micro stations are used to provide the required capacity under the coverage umbrella of macro stations Two types of BSs: q q n n Critical stations: usually the macrocell, which cannot be put into sleep mode, due to coverage issues Flexible stations: the BS that can be set in sleep mode No need to increase the cell ranges or the parameters of the BSs remaining on (low probability of coverage holes) Control schemes: n n Distributed approach: each flexible BS decides on the state of its operation (on/off), independently Pseudo-distributed approach: Flexible stations are assigned to critical stations, usually under cell overlap criteria Source: S. Kokkinogenis and G. Koutitas, “Dynamic and static base station management schemes for cellular networks,” in IEEE Global Communications Conference (GlobeComm ’12), Dec. 2012. Michela Meo – Politecnico di Torino Multi-tier network n Marco-micro network co-exists and cooperates with other technologies, e.g. femto cell and WiFi The main objective is to provide an offloading solution n With WiFi: n n n n Mobile operators usually do not have the administrative rights to the Wi-Fi APs Multi-Radio Access Technology (multi-RAT) is needed Limitations and constraints for integration with existing BS system q q Guarantee that coverage holes do not occur (especially indoor scenarios) Software and hardware limitations of real equipment (availability of low power states, transient times) Sources: • S. Kokkinogenis and G. Koutitas, “Dynamic and static base station management schemes for cellular networks,” in IEEE Global Communications Conference (GlobeComm ’12), Dec. 2012. • I. Haratcherev and A. Conte, “Practical energy-saving in 3g femtocells,” in IEEE Green Broadband Access (GBA) workshop, in conjunction with ICC 2013, Jun. 2013. • I. Haratcherev, M. Fiorito, and C. Balageas, “Low-power sleep mode and out-of-band wake-up for indoor access points,” in GLOBECOM Workshops, 2009 IEEE, 2009, pp. 1–6. Michela Meo – Politecnico di Torino Network sharing BS power The low load proportionality of the devices makes the whole access network little load proportional Network power n load load P = Pconst + f ( ρ ) Service provisioning cost is multiplied by the no. of networks Michela Meo – Politecnico di Torino Network sharing n Several competing mobile operators cover the same area with their equipment n During low traffic periods, when the resources of one (or a few) operator are sufficient to carry all the traffic, make the operator share their infrastructure q Switch off the network of one operator q Let users roam to other operators n In the short term, inter-operator switching schemes can reduce the waste In the long term, a unique efficient infrastructure with multiple virtual operators can be envisioned n Michela Meo – Politecnico di Torino Case study: Some European Countries Country France Germany Greece Italy Netherlands Poland Portugal Spain Romania Russia Ukraine U.K. MNOs 3 4 3 3 3 4 3 3 3 3 3 3 Market share [%] 46 36 19 32 31 21 16 51 28 21 38 36 26 46 26 28 29 29 28 14 45 40 15 44 34 22 41 32 26 37 33 30 48 37 15 39 33 28 - Subscr. [M] 58.2 113.6 15.4 84.0 19.0 47.5 16.4 51.4 24.2 189.7 52.3 68.5 Table 1: Characteristics of the considered countries: Number Michela Meo – Politecnico di Torino of MNOs offering both 2G and 3G services, market share for Case study: Some European Countries 0.8 consumer, const consumer, var business, const business, var 0.7 Relative saving 0.6 0.5 Large savings, more than 30%! 0.4 0.3 0.2 0.1 UK Ukraine Russia Romania Spain Portugal Poland Netherlands Italy Greece Germany France 0 Source: M. A. Marsan, M. Meo, “Network Sharing and its Energy Benefits: a Study of European Mobile Network Operators”, IEEE Globecom 2013 - Symposium on Selected Areas in Communications, December 2013. Michela Meo – Politecnico di Torino Case study: Some European Countries 1 consumer, WD consumer, WE business, WD business, WE The key point is increasing utilization 0.6 0.4 0.2 UK Ukraine Russia Romania Spain Portugal Poland Netherlands Italy Greece Germany France 0 No sharing Utilization 0.8 Michela Meo – Politecnico di Torino Mobile operator co-operation n Game-theory based model q q q q n Two MNOs cover the same area and offer the same QoS, but may have different network planning strategies Heterogeneous architecture is assumed (different BS types) Deployment targets peak of the user demand Binary transitions: all traffic is migrated from one to the other MNO The following game (transferrable utilities coalitional game of BS sharing) has been constructed: q q q MNO 1 has different strategies corresponding to the chosen roaming price MNO 2 has the choice between roaming or no roaming its traffic Game objectives: reduce own costs by cooperation, receive fair share of benefits of total cost reduction On-going work (iMinds, PoliTO, UTH, TUB): conference article “Greening the AirWaves: energyefficient BS sharing” under preparation. Michela Meo – Politecnico di Torino From energy efficient networking to sustainable networking n Power BSs with Renewable Energy Sources q q q Deploy cellular networks in countries where the power grid does not exist or is unreliable Achieve extremely low carbon footprints Survive natural disasters which damage the power grid Possible scenarios: 1. New opportunities for the deployments of networks in emerging regions Michela Meo – Politecnico di Torino Possible scenarios Michela Meo – Politecnico di Torino From energy efficient networking to sustainable networking n Power BSs with Renewable Energy Sources q q q Deploy cellular networks in countries where the power grid does not exist or is unreliable Achieve extremely low carbon footprints Survive natural disasters which damage the power grid Possible scenarios: 1. New opportunities for the deployments of networks in emerging regions 2. New business models (high electricity price, green incentives and sensitivity) Michela Meo – Politecnico di Torino Dimension PV powering for a BS n Consider a typical BS q q n n n Consumption profile Traffic profile energy need Choose a location Simulate energy production Simulate battery charge and discharge to decide system dimensioning Michela Meo – Politecnico di Torino 18 BS consumption Macro cell with LTE technology, with and without Remote Radio Unit (PA close to antenna) Power [W] Power [W] 1350 840 780 500 Psleep, 450 Psleep, 336 Deep sleep Deep sleep 0 • load 1 0 load 1 When needed (no TLC infrastructure) wireless backhauling consumes additional 200-250W, for a total of 30KWh/day Michela Meo – Politecnico di Torino 10 Traffic profiles: Business area 0 n n n 1 0.8 0.6 0.4 Parameter L 0.2 0 Fast Fig. transitions 4. Double switch-off: network saving versus the parameter L for the synthetic traffic profile. Peaks during the day Large difference weekday/weekend 1 Weekday Weekend Average Daily Consumption With RRU Without RRU Fig. 5. Traffic, f(t) 0.8 WD [KWh] WE [KWh] 15.2 12.55 23.9 19.5 0.6 0.4 0.2 0 00:00 05:00 Business cell: weekday and weekend traffic profiles. 10:00 15:00 Time, t [h] Michela Meo – Politecnico di Torino 20:00 00:00 20 Traffic profiles: Consumer n n n Slow transitions Peaks in the evening Little difference weekday/weekend 1 Weekday Weekend Average Daily Consumption Traffic, f(t) 0.8 WD [KWh] WE [KWh] With RRU 15.5 15.7 Without RRU 23.8 24.7 Fig. 6. 0.6 0.4 0.2 0 00:00 05:00 Consumer cell: weekday andMeo weekend traffic profiles.di Michela – Politecnico 10:00 15:00 Time, t [h] Torino 20:00 00:00 21 ! Figura 28: Radiazione solare annuale in Europa (Fonte: PVGIS, Institute for Energy - Unione Europea, 2012) Three locations Solar radiation Torino (4KWh/m2, high seasonal variance) Solar radiation ! Figura 29: Radiazione solare annuale in Africa (Fonte: PVGIS, Institute for Energy - Unione Europea, 2012) Assuan ! ! (6.8KWh/m2, low seasonal variance) ! 2) Palermo (5.1KWh/m H`! ura 28: Radiazione solare annuale in Europa (Fonte: PVGIS, Institute for Energy - Unione Europea, 2012) Michela Meo – Politecnico di Torino Dimensioning the system n Dimensioning based on the average of the days of the worst month: look for the system dimension that guarantees min. surplus ! Power [W] ! ! Figura 31: Produzione mensile di energia elettrica (Torino) Hour Figura 34: Potenza nel giorno di riferimento per Torino Location w/o RRU KWp w RRU KWp Torino 20 14 Palermo 16 10 Assuan ! 8 6 ! Figura 32 Produzione mensile di energia elettrica (Palermo) Michela Meo – Politecnico di Torino Storing energy: Battery dimensioning Energy waste Torino Battery charge [%] ! Low charge (spoils batteries) week Figura 43 Stato di carica delle batterie nell’arco di un anno –Torino – Impianto da 20 kWp Michela Meo – Politecnico di Torino ! Dimensioning the system: Panel area & batteries n Dimension battery so that charge never goes below 30% during the whole year Without RRU With RRU Location KWp Area [m2] No. batt. KWp Area [m2] No. batt Torino 20 98 75 14 68 45 Palermo 16 78 50 10 49 32 Aswan 8 39 30 6 29 16 Results strongly depend on the location Source: Marco Ajmone Marsan, Giuseppina Bucalo, Alfonso Di Caro, Michela Meo, Yi Zhang, Towards Zero Grid Electricity Networking: Powering BSs with Renewable Energy Sources”, IEEE ICC'13 - Workshop on Green Broadband access: energy efficient wireless and wired network solutions, June 2013. Michela Meo – Politecnico di Torino Dimensioning the system: Panel area & batteries n Dimension battery so that charge never goes below 30% during the whole year Without RRU With RRU Location KWp Area [m2] No. batt. KWp Area [m2] No. batt Torino 20 98 75 14 68 45 Palermo 16 78 50 10 49 32 Aswan 8 39 30 6 29 16 Very large! EE technology is needed Michela Meo – Politecnico di Torino 50(!0.&!A&/3!/)&)(*.@!5:3!/(!)4*E&!&!;3/)(43!0.&!A&//&!(.)3./()Q!2(!)4&88(5*@!'*//&!3.)4&43!(.!0.*! 50(!0.&!A&/3!/)&)(*.@!5:3!/(!)4*E&!&!;3/)(43!0.&!A&//&!(.)3./()Q!2(!)4&88(5*@!'*//&!3.)4&43!(.!0.*! /)&)*! 238(.()*! /+33'! 1*23C! T03/)*! +*!/+33'! /'3;.(13.)*! 2(! A0*.&! '&4)3! /)&)*!(1'+(5&! 238(.()*! 1*23C! T03/)*! (1'+(5&! +*!23++F:&42R&43! /'3;.(13.)*! 5:3! 2(! A0*.&! '&4)3! 23++F:&42R&43! 5:3! Combining RES with sleep modes 5*1'*.3!+&!A&/3!/)&)(*.@!(.!'&4)(5*+&43!+F&1'+(8(5&)*43!2(!'*)3.6&@!(.!1*2*!2&!)3.343!(!5*./01(! 5*1'*.3!+&!A&/3!/)&)(*.@!(.!'&4)(5*+&43!+F&1'+(8(5&)*43!2(!'*)3.6&@!(.!1*2*!2&!)3.343!(!5*./01(! '(S!A&//(C! '(S!A&//(C! n 2*E0)*! When traffic is below 50%, make 50% of the 8&4/(! 5&4(5*! /(&! 5*1'+3)&13.)3! /1&+)()*! 0.F&+)4&! A&/3! /)&)(*.! &! +3(! 2&! &2(&53.)3C! 2*E0)*! 8&4/(! 5&4(5*!2&! /(&! 5*1'+3)&13.)3! /1&+)()*! 0.F&+)4&!,.! A&/3! /)&)(*.! &! +3(! &2(&53.)3C! ,.! T0&.2*! 0.&! A&/3! /)&)(*.! ?! (.!T0&.2*! /+33'! 1*23! '43/0''*.3! (+! )4&88(5*! &E43AA3! 5:3! (+! )4&88(5*! 2(! 50(! &E43AA3! 0.&! /(! A&/3! /)&)(*.! ?! 5:3! (.! /+33'! 1*23!2(! /(!50(! '43/0''*.3! BS go to sleep mode 1*13.)*C! B03/)*! 1*2*! /(! 5*.)(.0&! &! ;&4&.)(43! +&! 5*1'+3)&! 2(/'*.(A(+()Q! 23+!+&! /(/)31&! (.! B0&+/(&/(! B03/)*! 1*2*! /(! 5*.)(.0&! &! ;&4&.)(43! 5*1'+3)&! 2(/'*.(A(+()Q! 23+! /(/)31&! (.! B0&+/(&/(! 1*13.)*C! T03/)*!/31'+(53!&+;*4()1*!2(!/+33'!1*23!?!4&''43/3.)&)*!(.!Y(;04&!_#C! T03/)*!/31'+(53!&+;*4()1*!2(!/+33'!1*23!?!4&''43/3.)&)*!(.!Y(;04&!_#C! ! Peak hours Figura 62 Figura 62 ! Off-peak hours U&!'*)3.6&!4(5:(3/)&!2&++&!A&/3!/)&)(*.!U9c!(.!/+33'!1*23@!5*13!;(Q!E(/)*!.3+!%&'()*+*!"@!?! U&!'*)3.6&!4(5:(3/)&!2&++&!A&/3!/)&)(*.!U9c!(.!/+33'!1*23@!5*13!;(Q!E(/)*!.3+!%&'()*+*!"@!?! 5*/)&.)3!3!'&4(!&-!! ! 5*/)&.)3!3!'&4(!&-!! ! !!" ! !!"# ! !!"##$ ! K"C"^L ! ! !!" ! !!"# ! !!"##$ K"C"^L ! Meo.3+! –2(!Politecnico di !Torino 2*E3! !!"# ! ?! (+! .0134*! 2(! &.)3..3! 4(53)4&/1())3.)(! /()*! 13.)43! ! !!"##$ ?! +&! '*)3.6&! 2*E3!Michela !!"# ! ?! (+! .0134*! &.)3..3! 4(53)4&/1())3.)(! .3+! /()*! 13.)43! ! !!"##$ ! ?! +&! '*)3.6&! 27 Use of sleep modes n While dimensioning of BS always on is basically unchanged, the dimensioning for BS going to sleep mode is much smaller No sleep With sleep Deep sleep Location Area [m2] Batt. Area [m2] Batt. Area [m2] Batt. Torino 98 75 78 52 39 27 Palermo 78 50 58 27 29 17 Assuan 39 30 29 18 14 13 Significant improvement! What about cost? Total cost: CAPEX+OPEX in 20 years Area type BS ISD [m] Coverage Grid only RES only [km2] [K€/km2] [K€/km2] Dense Urban 500 0.65 39.4 29.5 Urban 1000 2.60 9.8 7.4 Suburban 1732 7.79 3.3 2.5 Rural 4330 48.71 0.5 0.4 RES is cost effective! Michela Meo – Politecnico di Torino Hybrid powering: RES & grid Total cost: CAPEX+OPEX No RES Only RES Hybrid RES is cost effective! Michela Meo – Politecnico di Torino Conclusions n Higher degree of load proportionality is needed q n Load proportionality can be partially achieved through q q n To adapt to traffic variations BS sleep modes (intra-operator approach) Network sharing (inter-operator approach) New promising scenarios with RES powering q q q Sustainability Cost reduction But needs also EE solutions Michela Meo – Politecnico di Torino Thank you! Michela Meo – Politecnico di Torino Additional slides Michela Meo – Politecnico di Torino Mobile traffic explosion vs energy costs n Information and Communication Technology (ICT) sector is responsible for 2% of global CO2 emissions and consumes 2-10% of global energy n Mobile traffic explosion will further aggravate this picture: q n increase of BS/AP density to fulfill the demand Up to 70% of cellular operators OPEX are energy costs Radio access networks – prime target for energy saving Breakdown of energy consumption (one of the European mobile network operators) Source: S. Vadgama and M. Hunukumbure, Trends in Green Wireless Access Networks, 2011 Michela Meo – Politecnico di Torino Components of a Base Station (BS) n n Fixed power: cooling, main supply, DC-DC Load-proportional power: PA, BBU, RF Source: EARTH (Energy Aware Radio and network technologies) project, “Energy efficiency analysis of the reference systems, areas of improvements and target breakdown,” 2012. Michela Meo – Politecnico di Torino Base station power consumption n LTE is energy efficient: In urban areas, with a typical user density of 300 users/km2 q q q n With the same transmission power, an LTE macro BS covers q q n LTE à 18W/user WiMAX à 27W/user HSPA à 68Wer/user Urban area à about 0.22 km2 Suburban/rural environments à 2.6 km2 An LTE macro BS consumes around 1 kW q q Urban area à 4500W/km2 Suburban/rural environments à 400W/km2 Source: W. Vereecken, W. Van Heddeghem, M. Deruyck, B. Puype, B. Lannoo, W. Joseph, D. Colle, L. Martens, and P. Demeester, “Power consumption in telecommunication networks: overview and reduction strategies,” IEEE Communications Magazine, vol. 49, no. 6, pp. 62–69, 2011. Michela Meo – Politecnico di Torino Load proportionality and power saving When k-1 BSs out of k can be in sleep mode (ρ<1/k), the power per BS becomes: a + kbρ ( k −1) Psleep PS ( ρ ) = + k k 100 Normalized power consumption vs. load for different device types: n 10% proportion: the devices today n 50% proportion: the devices under development n 90% proportion: not realistic with today technologies 10% propor. Normalized consumption 50% propor. 10 90% propor. 1 n 0.1 0.001 with sleep no sleep 0.01 0.1 n 1 Sleep modes need to be introduced Need for BS management algorithms BS load, ρ Michela Meo – Politecnico di Torino Mobile operator co-operation n Main obstacles to network sharing: n n n MNOs are reluctant to allow their subscribers to frequently roam to a competitor MNOs QoS provisioning in the visited network (dominant vs small MNOs) Need for extended roaming and billing procedures Michela Meo – Politecnico di Torino ZEN: an example Orange green strategy for AMEA zone Project : optimized power consumption and solar powered mobile Morocco network Jordan Egypt Mali Niger Senegal Guinea Bissau Guinea Ivory Coast Cameroon Central African Republic Kenya Madagascar Equatorial Guinea + Dominican Republic, Vanuatu, Armenia, France the project was initiated in Africa and is now deployed in 18 countries of France Telecom-Orange fully integrated photovoltaïc solution Objectives sustainable rural Telecom network strong reduction of power consumption by selecting high efficiency telecom equipment, no active cooling smart power architecture and management – of Politecnico optimized solar energyMichela solution andMeo less use Diesel engines di sizing and techno-economic tools • 2065 network sites,development for 3.3 M people access greener solutions enabling Telecom in bad quality electric grid or off grid areas, • 13GWh solar energy produced in 2011 • 25 Mliters Innovationfuel 67 Ktons CO2 saved in 2011 Torino optimized and evolutive solution Dimensioning the PV system n A system is defined in terms of KWpeak: the achieved production in KW when radiation is 1KW/m2 Av. daily production [KWh] ! Torino month • Larger systems allow for larger production • Production changes according to seasons, while BS traffic and consumption not ! Figura 31: Produzione mensile di energia elettrica (Torino) Michela Meo – Politecnico di Torino Battery dimensioning Av. daily production [KWh] ! Torino To absorb variability more batteries are needed ! ! month Assuan Figura 31: Produzione mensile di energia elettrica (Torino) ! Michela Meo –Figura Politecnico Torino 33 Produzione di mensile di energia elettrica (Assuan)