Research on Energy Efficiency of 4G Cellular Networks with Co-channel Interference

Transcription

1 Research on Energy Efficiency of 4G Cellular Networks with Co-channel Interference rof. Xiaohu Ge & rof. Cheng-Xiang Wang Huazhong University of Science and Technology Heriot- Watt University

2 Outline Introduction System Model Spatial Distribution of Traffic Load Spatial Distribution of ower Consumption Energy Efficiency of VT Cellular Networks Conclusion A list of deliverables

3 Introduction(1) Energy efficiency issues in cellular networks (static analysis): The energy consumption of Mobile operator can be as high as 1 MW In EU: Transceiver Idling 19% 19 % 22 % ower Amplifier 22% ower Supply 16% 16 % 13 % 9 % 9 % 8 % 1 % 3 % Cabling 1% Transmit ower 3% Central Equipment 8% Cooling Fans 13% Combining/Duplexing 9% Transceiver ower Conversion 9% Over 8% of the power is consumed in RAN (at ChinaMobile) Base Station consumption kw BSs consume most energy in RAN

4 Introduction(2) Energy efficiency issues in cellular networks (dynamic): Temporal and spatial variations of traffic load in cellular networks Lasting exponential data traffic growth for at next five years, and more complicated behaviors that are shown to be self-similar and bursty In a dynamic cellular network, an energy efficiency model relating to traffic load variations is significant for dynamic energy-efficient BS planning, management, and operation.

5 Introduction(3) Importance of space in wireless networks, esp., energy consumption problems: TX-RX distance interference traffic load variations in space Tab. 1. Comparisons of time/frequency/space resources Resource Type TXs & RXs (spilling) ower falloff (interference) Time (division) Frequency (division) Space (division) collocated Collocated not collocated To zero at turn-off >=1dB/decade What about the impact of the spatial heterogeneity(hotspots) and randomness of traffic load towards energy efficiency in the interfering cellular networks? 2-4dB/decade J.G. Andrews, R.K. Ganti, M. Haenggi, etc. A rimer on Spatial Modeling and Analysis in Wireless Networks, IEEE Communication Magazine, vol. 48, pp , Nov. 21.

6 Introduction(4) Summary of current research in cellular energy efficiency Traffic-adaptive power management e.g.: Shutdown or sleep strategy, Macro- / micro- / femtocells overlaying, Adaptive traffic coalescing (ATC) neglect complex physical transmission processes, esp. under wireless channel effects and interference Energy-efficient transmission e.g.: energy-efficient power control / like adaption, MIMO/SIMO transmission mode switch limited in the link level of cellular networks To enable dynamic analysis, network level energy efficiency should be discussed by considering wireless channel effects, interference and traffic load characteristics.

7 System Model(1) - oisson-voronoi Tessellation Cellular Networks: BSs: y : k,1,2, ~oisson oint rocess ( B ), B Bk MSs: x : i,1,2, ~ ( M ); M Mi An MS is served by the nearest BS in range, which would suffer the least path loss during wireless transmission. The typical cell (alm theory) - Channel model: the channel gain of the link between BS k and its i-th user is c 2 Lki ( r,, ) rx tx K e r where Gaussian (,1) and the constant c ln1 / 1 ; the term exponentially distributed with mean 1 in Rayleigh fading environments. 2 is

8 km System Model(2) BS 5 BS 6 BS 4 BS 1 BS C MS BS2 BS 3 Fig. 1 Illustration of VT cellular structure; real lines depict cell boundaries inside which a polygon corresponds a cell coverage; dashed lines, which are perpendicularly bisected by corresponding cell boundaries, demonstrate how to build tessellations through the Delaunay Triangulation method km

9 System Model(3) where roblem formulation via the additive functional: def w i i x i M w x 1 x 2 wx : is a given non-negative function (either deterministic or random), 1... is an indicator function, and is a typical cell. Aggregate traffic load T in : w( x) ( x ) is the spatial traffic density. BS tranmission power in : w x ( x, I o, ), which is the power consumption (the energy cost ) of a typical point-to-point fading wireless link; I o represents other-cell interference and is the traffic density. Modeling work is needed. Empirical traffic characterization and modeling results provide us a basis. S.G. Foss, and S.A. Zuyev, On a Voronoi aggregative process related to a bivariate oisson process, Advances in Applied robability, vol. 28, no. 4, pp , Dec F. Baccelli, M. Klein, M. Lebourges, and S. Zuyev, Stochastic geometry and architecture of communication networks, Telecommunication Systems, vol. 7, no. 1, pp , 1997.

10 Spatial Distribution of Traffic Load(1) The aggregate traffic load in a typical cell is defined as: T def ( x ) 1 Mi Mi x Mi M x where ( x ) is the traffic intensity on each user, with DF and min f ( x), x 1 min x 1,2 reflects the heaviness of the distribution tail. The characteristic function of T Empirical measurement results have demonstrate that the traffic load in both wired and wireless networks, including cellular networks, is selfsimilar and bursty, which can be modeled by areto distributions with infinite variance. T j 1 ( j ) (, j ) M M min min b B b B a with (, ) 1 t j min t e dt j min

11 DF of aggregate traffic load Spatial Distribution of Traffic Load(2) erformance analysis of traffic load model 2.5 x 1-3 M / B =15 / M B =3 M / B = The probability mass (which can be depicted as the area under the DF curve) would shift to the right with the 1.5 increase of M B, indicating an increase in the average aggregate traffic load at BS Aggregate traffic load [kbps] Fig. 2. Aggregate traffic load in a typical VT cell with respect to the intensity ratio of MSs and BSs.

12 DF of aggregate traffic load Spatial Distribution of Traffic Load(3) erformance analysis of traffic load model 3.5 x =1.8, min =1 =1.8, min =15 =1.2, min =1 =1.2, min = The minimum traffic rate and heaviness index have inverse impacts on the aggregate traffic load in a typical VT cell Aggregate traffic load [kbps] Fig. 3 Impact of heaviness index and minimum traffic rate on the aggregate traffic load in a typical VT cell.

13 Spatial Distribution of ower Consumptions (1) Interference and power control model: The instantaneous SIR of MS is given by S S I Lr (,, ) agg S 1 ( L L ) k k k k D k kk k Lr ( kk, kk, kk) BS with 1D( Lk Lkk ) 1 D { rkk rk 1}, where I is the aggregate interference seen at agg MS, BS is the index set of interfering BSs, and the indicator function 1D( Lk Lkk ) is a constraint on MS distance distributions under the closest association rule in VT cellular networks.... IMS k IMS k+1 IBS k+1 Interfering downlinks r kk, ξ kk, ζ kk r k, ξ k, ζ k IBS k... MS Active downlinks Interfering BSs locate outside the dotted circle. r, ξ, ζ Fig. 4. Wireless downlinks of a VT cellular network. An example of interfering BS is illustrated at with detailed channel parameters. IBS k BS

14 Spatial Distribution of ower Consumptions (2) The required total transmission power in a typical VT cell with perfect power control: where Mj B _req K Mj x Mj M x y U 1 x c 2 U S V, V e. The characteristic function of i is: ( ) exp M 1 E _req B G( ) V K 2 2 B B with 2 G( ) G( ) 1 j sign( ) tan 4 Inf 2 2/ 2/ (1 )cos( ) E( Sk ) E( Qk ) B 2 E( Q ) exp, k sin(2 / ) 2/ 4c 2 2 2

15 Spatial Distribution of ower Consumption(3) The practical total transmission power of the typical BS limited to maximal power max, can be derived by truncating in the _req interval, max, f _pra ( x) f ( x) F ( ), x ; _req _req max, x ; A linear average BS power consumption model is built as follows E( ) E ( ) BS RF _pra max xf max f RF _req _req Circuit ( x) dx ( x) dx max Circuit where RF is the average efficiency of RF transmission circuits and the circuit power is fixed as a constant. Circuit max

16 DF of required total BS transmission power Spatial Distribution of ower Consumption(4) erformance analysis of BS power consumption.5.45 =1.9 =1.5 = Required total BS transmission power [W] With the decrease of heaviness index, indicating more bursty traffic load at MSs, the probability mass of required total BS transmission power remains rather stable except for the increasingly heavier tail that decays slower. Fig. 5. Required total BS transmission power with respect to heaviness index

17 DF of required total BS transmission power DF of required total BS transmission power Spatial Distribution of ower Consumptions (5) erformance analysis of BS power consumption M / B =15 M / B =3 M / B = Inf = Inf = Inf = Required total BS transmission power [W] Fig. 6. Required total BS transmission power with respect to the intensity ratio of MSs and BSs Required total BS transmission power [W] Fig. 7. Required total BS transmission power with respect to interfering link intensities

18 Energy Efficiency of VT Cellular Networks (1) Energy efficiency modeling: Energy efficiency metric: E(T ) 1 p EE M B C E( ) min BS out max _req ( x) dx max max ( 1) 1 RF ( ) _req Circuit ( ) _req QoS constraints: - Minimum data (traffic) rate x - BER target p b SIR gap - Maximal transmission power f xf x dx f x dx 2

19 Energy efficiency of VT cellular networks [bits/hz/joule] Energy Efficiency of VT Cellular Networks (2) Numerical results and discussions =1.2, min =2 =1.2, min =3 =1.8, min =2 =1.8, min = Intensity ratio between MSs and BSs EE,max M B opt.55,.45,.29,.26 bits/hz/joule 11, 8, 13, 9 The burstiness of traffic load causes the energy efficiency of VT cellular networks to fluctuate over a wide range. Fig. 8. Energy efficiency of VT cellular networks with respect to the intensity ratio of MSs and BSs considering the heaviness index and the minimum traffic rate.

20 Energy efficiency of VT cellular networks [bits/hz/joule] Energy Efficiency of VT Cellular Networks (3) Numerical results and discussions.4 Inf = Inf = Inf = EE,max M B opt.39,.29,.23 bits/hz/joule 17, 13, Fig. 9. Energy efficiency of VT cellular networks with respect to the intensity ratio of MSs and BSs considering the interfering link intensity Intensity ratio between MSs and BSs

21 Energy efficiency of VT cellular networks [bits/hz/joule] Energy Efficiency of VT Cellular Networks (4) Numerical results and discussions.5.45 =3.6 =3.8 =4 EE,max M B opt.17,.29,.46 bits/hz/joule 8, 13, To optimize energy efficiency, a tradeoff between the fixed and the dynamic BS power consumption in accordance with traffic load variations should be considered Intensity ratio between MSs and BSs Fig. 1. Energy efficiency of VT cellular networks with respect to the intensity ratio of MSs and BSs considering the path loss exponent.

22 Conclusion An energy efficiency model for oisson-voronoi tessellation (VT) cellular networks is proposed by considering spatial distributions of traffic load and power consumption. Simulation results have shown that there is a maximal limit of energy efficiency in VT cellular networks considering a tradeoff between the traffic load and BS power consumption. Moreover, wireless channel conditions have great impact on the energy efficiency of VT cellular networks. Our analysis indicates that interference reduction or interference coordination can effectively improve the energy efficiency of VT cellular networks, especially in scenarios with high intensity ratio of MSs and BSs.

23 A list of deliverables A. ublications [1] L. Xiang, Xiaohu Ge (corresponding author), Cheng-Xiang Wang, Frank Y. Li and Frank Reichert, Energy Efficiency Evaluation of Cellular Networks Based on Spatial Distributions of Traffic Load and ower Consumption, IEEE Trans. On Wireless Commun., minor revision. [2] Xiaohu Ge, K. Huang, Cheng-Xiang Wang, X. Hong, Capacity Analysis of a Multi-Cell Multi-Antenna Cooperative Cellular Network with Co-Channel Interference, IEEE Trans. On Wireless Commun., vol. 1, no. 1, pp , Oct [3] I. Humar, Xiaohu Ge (corresponding author), L. Xiang, J. Ho, M. Chen, Rethinking Energy Efficiency Models of Cellular Networks with Embodied Energy, IEEE Network Magazine, Vol.25, No.3, pp.4-49, March, 211. [4] L. Xiang, Xiaohu Ge, C. Liu, L. Shu, Cheng-Xiang Wang, A New Hybrid Network Traffic rediction Method, IEEE roc. Conf. GlobeCom 21, Miami, USA, Dec. 21. (A Best aper Award) [5] Z. Chen, Cheng-Xiang Wang, X. Hong, J. S. Thompson, S. A. Vorobyov, Xiaohu Ge, H. Xiao, and F. Zhao, Aggregate interference modeling in cognitive radio networks with power and contention control, IEEE Trans. Commun., Vol.6, No.2, pp , Feb., 212. [6] X. Hong, Cheng-Xiang Wang, J. S. Thompson, B. Allen, W. Q. Malik, and Xiaohu Ge, On space-frequency correlation of UWB MIMO channels, IEEE Trans. on Vehicular Technology., vol. 59, no. 9, pp , Nov. 21. [7] X. Hong, Cheng-Xiang Wang, M. Uysal, Xiaohu Ge, and S. Ouyang Capacity analysis of hybrid cognitive radio networks with distributed VAAs, IEEE Trans. Vehicular Technology, vol. 59, no. 7, pp , Sept. 21.

24 [8] Cheng-Xiang Wang, X. Hong, Xiaohu Ge, X. Cheng, G. Zhang, and J. S. Thompson, Cooperative MIMO channel models: a survey, IEEE Communications Magazine, vol. 48, no. 2, pp. 8-87, Feb. 21. [9] Xiaohu Ge, J. hu, Cheng-Xiang Wang, J. Zhang and X. Yang Energy Efficiency Analysis of MISO-OFDM Communication Systems Considering ower and Capacity Constraints, ACM Mobile Networks and Applications, Vol.17, No.1, pp.29-35, Feb [1] Xiaohu Ge, Cheng-Xiang Wang, Y. Yang, L. Shu, C. Liu, and L. Xiang, AFSO: an adaptive frame size optimization mechanism for IEEE wireless networks, KSII Trans. on Internet and Information Systems, vol. 4, no. 3, pp , June 21. [11] Xiaohu Ge, Y. Yang, Cheng-Xiang Wang, Y.-Z. Liu, C. Liu, and L. Xiang, Characteristics analysis and modeling of frame traffic in wireless networks, Wireless Communications and Mobile Computing, John Wiley & Sons, vol. 1, no. 4, pp , Apr. 21. [12] Z. Chen, C.-X. Wang, X. Hong, J. S. Thompson, S. Vorobyov, F. Zhao, and Xiaohu Ge, Interference mitigation for cognitive radio MIMO systems based on practical precoding, Invited aper, Elsevier hysical Communication, accepted for publication. [13] F. S. Haider, Cheng-Xiang Wang, H. Haas, E. Hepsaydir, and Xiaohu Ge, Energy-efficient subcarrier-and-bit allocation in multi-user OFDMA systems, in roc. IEEE VTC 12-Spring, Yokohama, Japan, May 212. [14] Y. Yuan, X. Cheng, Cheng-Xiang Wang, D. I. Laurenson, Xiaohu Ge, and F. Zhao, Space-time correlation properties of a 3D two-sphere model for non-isotropic MIMO mobile-to-mobile channels, roc. IEEE Globecom 1, Miami, USA, Dec. 21. [15] X. Cheng, Cheng-Xiang Wang, Y. Yuan, D. I. Laurenson, and Xiaohu Ge, A novel 3D regular-shaped geometrybased stochastic model for non-isotropic MIMO mobile-to-mobile channels, invited paper, roc. IEEE VTC 1-Fall, Ottawa, Canada, 6-9 Sept. 21. [16] Z. Chen, Cheng-Xiang Wang, X. Hong, J. S. Thompson, S. A. Vorobyov, and Xiaohu Ge, Interference modeling for cognitive radio networks with power and contention control, roc. IEEE WCNC 21, Sydney, Australia, Apr. 21. [17]A. Ghazal, Cheng-Xiang Wang, H. Haas, M. Beach, X. Lu, D. Yuan, and Xiaohu Ge, A Non-Stationary MIMO Channel Model for High-Speed Train Communication Systems, roc. IEEE VTC 212-spring, May 6-9, Yokohama, Japan.

25 B. Joint rojects 1. China Hubei rovincial Science and Technology Department, Joint Research on Key Technologies of Next Generation Green Broadband Mobile Communications, rincipal Investigator: Xiaohu Ge, Foreign artner: Cheng-Xiang Wang, Research eriod: 211~213, Budget: 1, RMB (1,GB) NSFC (National Natural Science Foundation China) Major International Joint Research roject, Research on Theory and Key Technologies of Information Spatial Cooperation Optimization in Green Communication Networks, Joint submitted by Xiaohu Ge and Cheng- Xiang Wang.

26 C. Research latform Green International Collaboration Research Base Green broadband wireless mobile communication (GREEN) Lab Granted by Hubei rovincial Science and Technology Department

27