SCIENTIA SINICA Informationis, Volume 47 , Issue 6 : 771-788(2017) https://doi.org/10.1360/N112017-00033

Energy-delay tradeoff and optimal base station sleeping control in hyper-cellular networks

More info
  • ReceivedFeb 15, 2017
  • AcceptedMar 21, 2017
  • PublishedJun 9, 2017


By decoupling control coverage and traffic coverage, dynamic base station (BS) sleeping becomes possible in hyper-cellular networks. Traffic coverage can be adjusted based on real-time demands, to enhance the energy efficiency of the networks. However, sleeping mechanisms may affect the networks' delay performance. Thus, it is important to obtain analytical results on the influence of sleeping mechanisms on the system's energy efficiency and delay performance. The optimal sleeping mechanism can be derived based on the energy-delay tradeoff. This work summarizes the relevant contributions of our group: based on the vacation queueing model, closed-form expressions for average power and mean traffic delay were obtained. The conditions for the energy-delay tradeoff were analyzed. Further, the optimal BS sleeping design and the optimal energy-delay tradeoff were obtained, assuming fixed delays. The influence of the mode-changing energy cost was studied. BS sleeping designs for bursty traffic were analyzed, combined with power matching. The optimal BS sleeping mechanism for bursty traffic was obtained with a partially observable Markov decision process (POMDP).

Funded by



  • Figure 1

    BS operation transition diagram for the single sleep (SS) scheme

  • Figure 2

    For the multiple sleep (MS) scheme, the normalized average power consumption vs. average delay when the setup time changes. $h_D=0$ s, $h_V=10$ s

  • Figure 3

    For the $N$-limited scheme, average delay vs. $N$. $h_S=3$ s, exponentially distributed hysteresis time with $h_D=1$ s

  • Figure 4

    (Color online) For the single sleep scheme, the normalized average power consumption vs. average delay when hysteresis time changes. Exponentially distributed setup time with $h_S=5$ s

  • Figure 5

    (Color online) Optimal power consumption (normalized by the case without sleep mode) vs. average delay constraint for different schemes. $c_S^2=0$, $h_S=0.25$ s

  • Figure 8

    State transition diagram for the extended IPP/$M$/1 queueing model with the$N$-based BS sleeping mechanism and power matching

  • Figure 11

    (Color online) Optimal sleeping threshold $N^*$ and transmit power $P_\text{TR}^*$ under IPP traffic with different arrival rate. $\tau=2$ s, $k=1$, $l=2$ MB

  • Figure 12

    (Color online) Total power consumption and average delay vs. system utilization $\rho$ and autocorrelation coefficient $\theta$, with SPP traffic $C^2=10$. $r_1= r_2$, $l=2$ MB, $N=1$, $P_\text{TR}=10$ W. (a) Total power consumption; (b) average delay

  • Figure 13

    (Color online) Optimal BS sleeping mechanism with varying BS current mode under IBP traffic. $E_\text{active}=$protectłinebreak 25 J, $\omega=10$, $\rho=0.2$. (a) $E_\text{SW}=417$ J, arrival state is OFF; (b) $E_\text{SW}=417$ J, arrival state is ON

  • Figure 14

    (Color online) System cost with varying traffic burstiness under different weight $\omega$. $E_\text{SW}=41.7$ J, $E_\text{active}=\\25$ J, $\rho=0.2$

Copyright 2020  CHINA SCIENCE PUBLISHING & MEDIA LTD.  中国科技出版传媒股份有限公司  版权所有

京ICP备14028887号-23       京公网安备11010102003388号