Toward 5G:A Novel Sleeping Strategy for Green Distributed Base Stations in Small Cell Networks

Transcription

1 th International Conference on Mobile Ad-Hoc and Sensor Networks Toward 5G:A Novel Sleeping Strategy for Green Distributed Base Stations in Small Cell Networks Yiwei Xu, Jin Chen, Ducheng Wu and Wanru Xu College of Communications Engineering, PLA University of Science and Technology, China Abstract Confronted with the rapidly increasing demand of mobile traffic and heavy energy consumption on base stations (BSs), the base station (BS) sleeping strategy becomes a promising method to promote the system energy efficiency (EE). To switch off the redundant small-cell base stations (s-bss) without whittling down the network EE in small cell networks, we propose a novel environment-friendly distributed BS sleeping strategy, which consists of two parts, matching and connecting between the s-bss and the user equipments (UEs), and activating the BS-turning-off procedure for the further promotion of EE. An initializing matching connection algorithm (IMCA) is proposed for the first subproblem and the energy efficiency ratio (EER) obtained by this proposed IMCA surpasses the EE gained from traditional initializing random connection method. Moreover, a turn-off if possible algorithm (TIPA) is proposed as the sleeping strategy to handle the BS-turning-off procedure. Both the EE and the convergence speed obtained by the sleeping deployment using TIPA surpass those of the traditional best response algorithm. I. INTRODUCTION A. Motivations Face up to the coming 5G era, millions more base stations (BSs) with higher functionality and billions more smart devices with much higher data rates will be running [1]. All these phenomena inevitably incur massive energy consumption [2], [3]. As a result, for the environmental protection [4], reducing the energy consumption of cellular networks has attracted increasing attention recently [5]. Accordingly, the environment-friendly base station (BS) deployment is expected to be a key design for the next generation mobile communication systems (5G) [6]. On the other hand, since the electricity bill of the entire cellular networks mainly comes from energy consumption on BSs [8], BS sleeping strategy has been recognized as one of the most effective energy efficient technologies [5]. Therefore, we will investigate the BS sleeping strategy in small cell networks (SCNs) under a self-organizing scheme. B. Related Works The existing literatures [9], [10], [12] focused univocally on the related topics. All these existing works [9], [10], [12] have been ignoring the actual preference of the UEs, so they mostly use the stochastic geometry approach and the random selection method to build the topological connections between UEs and BSs, without taking the preference of the UEs into account. The optimal combination of macro and micro BS densities is studied in [9], where the energy consumption is significantly reduced by sleeping control. Whereas, [9] only optimized the energy consumption without considering the energy efficiency (EE), which leads to less satisfaction of the user equipments (UEs). And the energy-efficient scheme to deploy and plan small cells according to the prevailing traffic pattern [12]. However, the BS deployment result obtained in [12] can hardly adjust with time the addition and subtraction of the changing UEs once they were settled. However, the scale and distribution of UEs are randomly or regularly changing, one fixed BS deployment can hardly satisfy all user patterns. Aim at above puzzles, we propose an energy efficient sleeping strategy for BSs in small cell network based on matching game. C. Challenges and Contributions There are three major challenges of our proposed selforganizing BSs sleeping strategy. First, our work considers the preference of the UEs and brings the matching games into sleeping strategy, which is complicated when two sides are engaged in choosing the other side. Second, as the conventional matching approach will only do the matching once [13] [15], it is very challenging to circulate the matching process. Third, the proposed sleeping strategy is self-organizing and distributed, where the BSs can decide their working mode by themselves. This paper provides a novel perspective in self-organizing BSs sleeping strategy to promote the network EE without compromising the UE throughput satisfaction. It is assumed that one BS can serve limited measurable UEs. Under this circumstances, the initialization can be modeled as a manyto-one matching market [11]. After the matching initialization, BSs will be randomly chosen to switch off if the UEs served by the chosen BS can all be assigned to other BSs. The main contributions of this paper can be summarized as follows: It is the first work of introducing matching theory into the sleeping strategy deployment in small cell network, instead of random associations. We propose an initializing matching connection algorithm (IMCA) and a novel self-organizing BSs sleeping strategy, called turn-off if possible algorithm (TIPA). The EE ratio of the proposed IMCA surpasses the random connecting scheme. The proposed TIPA improves the EE without compromising the UEs throughput, and the performance of TIPA also surpasses the traditional best response algorithm. Moreover, the convergence speed /16 $ IEEE DOI /MSN

2 of the proposed TIPA is faster than the best response algorithm. The rest of this article is organized as follows. In Sec. II, we present the system model and formulate the problem of BSs deployment based on matching game theory. In Sec. III, we propose an initializing matching connection algorithm (IMCA) to solve the assignment problem and a turn-off if possible algorithm (TIPA) to address the deployment. In Sec. IV, the simulation results and the discussion are presented. Finally, we make conclusions in Sec. V. II. SYSTEM MODEL AND PROBLEM FORMULATION A. System Model In this sleeping strategy, we are about to design, assume that every s-bs has two working modes, active mode and sleeping mode, between which they can switch freely. When the distribution of UEs changes, some s-bss may decide to switch their mode, and the network connecting topology between the s-bss and the UEs may reschedule. Under a certain distribution of UEs, the redundant s-bss tend to switch off as long as the UEs served by them can all be reconnected to the other s-bss. It is also assumed that every s-bs has a maximum serving UE number and when the number of the connected UEs to one certain s-bs reaches the maximum serving threshold, the s-bs will know the overflow and trigger the sorting arrangement as long as a preferential selection. Assume that all the s-bss are connected with a virtual central controller as well as a data center, so every s-bs is aware of the service state (active or sleeping) and the remaining service place of other s-bss. We consider deploying an SCN in the urban area with N candidate small cell base stations (s-bss) N = {n 1,n 2,..., n N } and M randomly distributed UEs M = {m 1,m 2,..., m M } as depicted in Fig. 1. The locations of s- BSs are selected uniformly in order to meet the demand of random uniformly distributed UEs. To satisfy all the UEs, the locations of s-bss can be arranged as dense as possible. However, the more s-bss we place, the more basic consumption will cost, even worse, the deployment of s-bss stays settled once they were deployed. Thus, it is urgent and indispensable to mitigate the power and energy consumption by s-bss sleeping strategy. Under a certain distribution of UEs, the state of s-bss are denoted by S =[s i ] 1 N, which represents if s-bs i is active (s i =1) or sleeping (s i =0). While L =[l i,k ] N M presents the connection state between s-bs i and UE k, if the UE k is served by s-bs i there is l i,k = 1, otherwise l i,k = 0. Assume that every UE will be served only by one s-bs, that N is l j,k =0. The signal interference noise ratio (SINR) j=1,i =j from s-bs i to UE k is given by γ i,k = p i,kg i,k, (1) σ where p i,k denotes the transmission power allocated to UE k by s-bs i, g i,k is the channel gain between s-bs i and UE k, Fig. 1: The system model before initialization. and σ is the power of UE terminal noise. The channel gain is defined by g i,k = d α i,k, where d i,k is the distance between s-bs i and UE k, α is the attenuation factor. Accordingly, it is assumed that UE k can be served by the s-bs i when γ i,k exceeds a certain threshold. Thus, the Shannon capacity gained by UE k is obtained as R i,k =log 2 (1 + γ i,k ), where γ i,k is the SINR given earlier. The communication consumption will be alleviated by switching off s-bss, on the basis of ensuring all the UEs being well served. Although the s-bss consume very limited power and energy during the sleeping mode, we can t turn the s-bss off blindly without considering the UEs satisfaction. Therefore, to measure the effectiveness of the SCNs system, the system energy efficiency ratio (EER) is designed as the ratio of the total throughput and the total energy consumption, modified from [12], which is N M S i l i,k R i,k i=1 k=1 η total = [ ], (2) N M S i (p i,k l i,k + Pi A)+(1 S i)pi S i=1 k=1 where P A i and P S i are the circuit power consumption of s-bs i when it is active and sleeping, respectively. The decision of s-bss is not only about switching on or off, but also about serving which UEs that will promote the EER. That is to say, the s-bss will judge and make the decision about the active or sleeping mode according to the distribution of UEs. B. Problem Formulation To switch off the redundant s-bss without whittling down the network EE, the proper initializing connections between s-bss and UEs are fundamental. Thus, the entire puzzle is divided into two subproblems, one is the optimal stable matching solution between s-bss and UEs, the other is the tentative closing procedure of redundant s-bss and the optimal s-bss sleeping-active deployment profile to achieve the maximization of EE. 116

3 1) The matching and connecting: The first subproblem is to determine the initial connections between s-bss and UEs as the Function. 3 shown. To realize this goal, we build the problem as the many-to-one matching game model, where s- BSs and UEs are matched to each other as two sides of the market. P 1 : G(N, M, n, m,θ(m),θ(n )) (3) where the s-bss N = {n 1,n 2,..., n N } and the UEs M = {m 1,m 2,..., m M } are denoted in the above, and N M=. Two preference relations are signified by n and m, θ(m) is the preference lists of the UEs, identically, θ(n ) is the preference lists of the s-bss. Suppose that both the s-bss and the UEs decide their preference by SINR, the goal of our matching game model is to search the optimal stable matching solution between s-bss and UEs under a certain network topology. 2) s-bss sleeping strategy for EE lifting: The second subproblem of this article is to find the optimal s-bss sleepingactive deployment profile to achieve the maximization of EE, which is presented as the following Function. 4. Dealing with task P 2, we propose a novel tentative method, which switches off the active s-bss as long as the UEs can still be served by other active s-bss. P 2 : max(η total ) (4) Different from the traditional best response approach, where the s-bss decide according to the EE calculated every time, in our proposed approach, the s-bss only need to judge the feasibility before turning off every time. In this way, the EE calculation overhead and time cost before every decision making are remarkably mitigated. III. THE PROPOSED DISTRIBUTED BSS SLEEPING STRATEGY Based on the matching game theory we propose an initializing matching connection algorithm (IMCA) for Subproblem. 1, which used to addressed by initializing random connection algorithm (IRCA). And we also propose a novel distributed self-organizing turn-off if possible algorithm (TIPA) for energy efficiency (EE) optimization in Subproblem. 2, which is settled by the best response algorithm. Both the proposed IMCA and TIPA will be presented elaborated in details in the following subsections. A. The Proposed Initializing matching connection algorithm (IMCA) To tackle the first subproblem, the proposed initializing algorithm of active-sleeping deployment based on the matching game theory, named initializing matching connection algorithm (IMCA) is described below, which mainly consists of three stages: preference lists generation, matching evaluation, rearrangement of still-not-served UEs. Instead of the randomly connection scheme in [12], which we call the initializing random connection algorithm (IRCA) in the Section. IV for Algorithm 1 The proposed Turn-off If Possible Algorithm (TIPA) Input: s-bss N,UEs M,coordinates of s-bss and UEs, the connection result. Output: The sleeping strategy of s-bss and the new connection after deployment. 1: for iteration do 2: for i =0to N do 3: Randomly choose one s-bs i to turn off. Reschedule the UEs once served by the chosen s-bs i. 4: for m {a, l i,a =1} do 5: Obtain the next option s-bs n of UE m from its preference sequence set θ(m) =γ i,m. 6: if P (n) =M max M l n,k > 0 then 7: k=1 if γ i,m > Γ then 8: l n,m =1. break 9: end if 10: end if 11: end for 12: if UEs m {a, l i,a =1} are successfully rearranged. then 13: S i =0. 14: for m {a, l i,a =1} do 15: l i,m =0. 16: end for 17: else 18: Cancel the off process and recover the settings. 19: end if 20: end for 21: end for convenience, our proposed IMCA comprehensively considers the preference of both the s-bss and the UEs. Stage 1 Calculate the SINR γ i,k between s-bs i and UE k using Func. 1. Sort the UE i s preference by the calculated results, and collect the UEs whose first choice is s-bs i, connect the UEs with their first choice s-bss if the SINR and the serving place is available. Stage 2 Connect each s-bs to the top-ranked UEs if available and the UEs are within the serving range of the s-bs. Move the UEs beyond the threshold of the serving place to the still-not-served set. Stage 3 Rearrange the UEs in the still-not-served set. B. The Proposed Turn-off If Possible Algorithm (TIPA) In the traditional best response approach, every time before turning off one s-bs, except for the predicating of the success of rearrangement, the system also needs to calculate the system 117

4 Fig. 4: The network EER before and after sleeping deployment under the same UEs-Density. Fig. 2: The initializing topology with 16 candidate s-bss and 100 randomly distributed UEs. Fig. 5: The network EER before and after deployment under different UEs-Density. Fig. 3: The sleeping deployment result after the proposed TIPA, with 16 candidate s-bss and 100 randomly distributed UEs. EE in case overdoes of switching off s-bss. However, the overhead cost and time consumption accumulate over and over again inevitably and turn out to be extremely heavy. In order to solve the second subproblem in a green way, the proposed turn-off if possible algorithm (TIPA) is presented in chart Algorithm. 1. First, randomly choose one s-bss and trigger the matching process to rearrange the UEs used to be connected to the chosen s-bs. Next, decide whether all the UEs served once by the chosen s-bs, can all be rescheduled to other s-bss. If the answer is yes, then we can switch off the chosen s-bs. Otherwise, loop to choose another s-bs randomly and repeat the following procedure. Until there is no more s-bs can be switched off, the final s-bs sleeping-active deployment is obtained. Fig. 6: Comparison of the Convergence of two algorithms. IV. SIMULATION In the area of 160m 160m, the 16 candidate s-bss locations are uniformly distributed with 100 randomly distributed UEs. Partly referenced [12], the transmission power p i,k is settled to 16 dbm, the circuit power consumption of active s- BS is 39 dbm, and the circuit power consumption of sleeping s-bs is 35 dbm. 118

5 A. The connection topology after sleeping strategy First of all, we use the proposed initializing matching connection algorithm (IMCA) to create the initial connection topology between the s-bss and the UEs. As illustrated in Fig. 2, the initializing matching connection topology after the proposed IMCA, with randomly distributed uniform 100 UEs. Before the deployment presented in Sec. IV-B, the UEs are all matched and connected with the corresponding s-bss nearby. Nevertheless, it is more or less extravagant for the active s-bss to serve only a few UEs, far less than the serving threshold, which mediately leads to low system EE. Therefore, it is desperately urgent to turn part of not-fullyutilized s-bss off. Even if the UE distribution is denser, a few less-loaded s-bss can still be switch off by both the best response algorithm and our proposed turn-off if possible algorithm (TIPA), Algorithm. 1. B. Comparison of the best response and the proposed TIPA In this subsection, the properties of the traditional best response algorithm and the proposed TIPA will be discussed. First, after switching off as much as possible BSs without compromising the system throughput, there are 5 s-bss turnedoff among the total 16 s-bss. As depicted in Fig. 3, the redundant s-bss, marked with the green triangles, are turned off. While the necessary s-bss, marked with the blue filled triangles, are still active operating. Next, we investigate the EER in two scenarios. One is under the same UE density with the changing serving area and the accordingly changing s-bss number. We compare the EER before and after the sleeping deployment, meanwhile, we also compare the EER after the traditional best response algorithm and the proposed TIPA. As presented in the following Fig. 4, the EER with only initial connecting is low. It is also obvious that after sleeping deployment, both the traditional best response approach and the proposed TIPA approach have a great promotion on EER. Better still, our proposed TIPA surpassed the traditional best response algorithm as the serving area grows. The second scenario is the under the same serving area with different UE density, as illustrated in Fig. 5. It is delighted to see that the EER obtained from the proposed IMCA plus the proposed TIPA surpasses the EER gained from the combination of the proposed IMCA plus the best response. Moreover, the EER of our proposed initial connecting method IMCA is far better than the traditional IRCA. Afterwards, the convergence of the proposed TIPA is faster than the best response algorithm as illustrated in Fig. 6. V. CONCLUSION To switch off the redundant s-bss without whittling down the network EE in the small cell networks, we have proposed a novel environment-friendly distributed BS sleeping strategy, which consists of two parts, one is the initial matching and connecting between the s-bss and the UEs, the other is the BS-turning-off procedure. An initializing matching connection algorithm (IMCA) is proposed for the first subproblem and the EER obtained by this IMCA surpasses the EER gained from traditional random connection method. Moreover, a turnoff if possible algorithm (TIPA) is proposed to tackle the subproblem of BS-turning-off. Both the EER and the convergence speed obtained by the sleeping deployment using TIPA surpasses that of the best response algorithm. ACKNOWLEDGMENT This work is supported by the Natural Science Foundation for Distinguished Young Scholars of Jiangsu Province under Grant No. BK , the National Natural Science Foundation of China under Grant No and No , and in part by the Open Research Foundation of National Key Laboratory of Science and Technology on Information Transmission and Dissemination Technologies in Communication Networks. REFERENCES [1] C. L. I, C. Rowell, S. Han, Z. Xu, G. Li and Z. Pan, Toward green and soft: a 5G perspective, IEEE Communications Magazine., vol. 52, no. 2, pp , February [2] M. A. Marsan and M. Meo, Network sharing and its energy benefits: A study of European mobile network operators, Proc. IEEE GLOBECOM, 2013, pp [3] [3] H. Zhang, A. Gladisch, M. Pickavet, Z. Tao, and W. Mohr, Energy efficiency in communications, IEEE Commun. Mag., vol. 48, no. 11,pp , Nov [4] K. Son, H. Kim, Y. Yi, and B. Krishnamachari, Base station operation and user association mechanisms for energy-delay tradeoffs in green cellular networks, IEEE J. Sel. Areas Commun., vol. 29, no. 8, pp , Sep [5] D. Feng, C. Jiang, G. Lim, L. J. Cimini, Jr., G. Feng, and G. Y. Li, A Survey of Energy-Efficient Wireless Communications, IEEE Commun. Surv. Tutorials, vol. 15, no. 1, pp , [6] J. G. Andrews, S. Buzzi, W. Choi, S. V. Hanly, A. Lozano, A. C. K. Soong, and J. C. Zhang, What Will 5G Be?, IEEE J. Sel. AreasCommun., vol. 32, no. 6, pp , Jun [7] E. Oh, B. Krishnamachari, X. Liu, and Z. Niu, Toward dynamic energy efficient operation of cellular network infrastructure, IEEE Commun. Mag., vol. 49, no. 6, pp , Jun [8] G. Auer et al., How much energy is needed to run a wireless network?,ieee Wireless Commun., vol. 18, no. 5, pp , Oct [9] D. Cao, S. Zhou and Z. Niu, Optimal Combination of Base Station Densities for Energy-Efficient Two-Tier Heterogeneous Cellular Networks, IEEE Trans. Wireless. Commun., vol. 12, no. 9, pp , Sep [10] Y. S. Soh, T. Q. S. Quek, M. Kountouris and H. Shin, Energy Efficient Heterogeneous Cellular Networks, IEEE J. Sel. Areas Commun., vol. 31, no. 5, pp , May [11] A. E. Roth and M. A. O. Sotomayor, Two-sided matching: A study in game-theoretic modeling and analysis, Cambridge University Press., [12] L. Zhou, Z. Sheng, and L. Wei, Green cell planning and deployment for small cell networks in smart cities, AD HOC NETWORKS, JUN [13] Jorswieck AE, Stable matchings for resource allocation in wireless networks, 17th International Conference on Digital Signal Processing (DSP), Corfu, Greece, 2011; 1-8. [14] Semiari O, Saad W, Valentin S, Bennis M, Maham B, Matching theory for priority-based cell association in the downlink of wireless small cell networks, International Conference on Acoustics, Speech and Signal Processing (ICASSP), Florence, Italy, 2014; 444C448. [15] Pantisano F, Bennis M, Saad W, Valentin S, Debbah M, Matching with externalities for context-aware user-cell association in small cell networks, Global Communications Conference (GLOBECOM), Atlanta, GA, 2013;