Joint Dimensioning of Outdoor Heterogeneous Radio Access Networks (HetNet) using Monte Carlo Simulation

Transcription

1 MACRo th International Conference on Recent Achievements in Mechatronics, Automation, Computer Science and Robotics Joint Dimensioning of Outdoor Heterogeneous Radio Access Networks (HetNet) using Monte Carlo Simulation Péter RATKÓCZY 1, Attila MITCSENKOV 1 Department of Telecommunications and Media Informatics, Faculty of Electrical Engineering and Informatics, Budapest University of Technology and Economics, Hungary ratkoczy@tmit.bme.hu, mitcsenkov@tmit.bme.hu Manuscript received January 25, 2015, revised February 9, Abstract: The experienced mobile traffic increase in the recent years made traffic capacity the bottleneck instead of the coverage constraints, calling for significantly higher density of the base stations. Heterogeneous radio access networks (HetNet) provide a possible solution to this problem, combining various wireless technologies. In this paper we investigated the joint dimensioning of the co-existent radio access networks, the relation between the required macro and small cell densities to meet a certain traffic demand, and compared the two main, competing technological solutions, namely small cells and Wi-Fi, suitable to complement an LTE (macrocell) network. Keywords: LTE, small cell, Wi-Fi, mobile, dimensioning, Monte Carlo, simulation, heterogeneous radio access network, HetNet 1. Introduction Due to the ongoing (and upcoming) changes in user behaviour in mobile networks, namely a significant shift towards highly traffic-intensive applications, the mobile networks are facing tough challenges in terms of traffic/capacity requirements. Heterogeneous mobile networks (HetNet) provide a possible way to handle the mobile traffic increase [1][2]. HetNet means the concurrent use of multiple types of wireless technology, e.g. macrocells, small cells, or Wi-Fi; with a complex cooperation between them. The main difference between pico cells and Wi-Fi is that pico base stations work in the licensed spectrum, which may lead to interference between an LTE macro base station and LTE pico base stations [3][4]. On the other hand, Wi-Fi works in the ISM band without interfering the LTE networks, but faces the physical limitations of the ISM band [5] /macro

2 136 P. Ratkóczy, A. Mitcsenkov The aim of this paper is to investigate the joint dimensioning of a heterogeneous network, when the small cells or Wi-Fi APs (Access Point) are part of the mobile network dimensioning process, and not taken as independent add-ons to the macrocells. Such joint dimensioning brings a more efficient use of components of the heterogeneous network, and reveals various combinations of macro and small cells, or macro cells and Wi-Fi APs. For the sake of such joint dimensioning, a software module was developed, which takes into consideration the technology specific radio propagation models, the interference in the network and the characteristics of the physical area (dense urban, urban, rural). The traffic capacity requirements to be met by the radio network are fine-tuned by several parameters (user s density, the active vs. heavy user ratio and their capacity demand). The paper is organized as follows: Section 2 describes the dimensioning process, including the model applied to derive radio bandwidth from traffic requirements, and calculation of the antenna densities, Section 3 is devoted to the numerical results of the dimensioning, while Section 4 concludes the paper and presents our main findings about outdoor Heterogeneous Networks. 2. Joint dimensioning of Heterogeneous Networks Calculating the necessary radio bandwidth with respect to the user s traffic demand (bitrate) is a fundamental element of the dimensioning. The relation between the radio bandwidth and the bitrate is the spectral efficiency (SE), which is a function of the signal quality, namely the signal-to interference-plus noise ratio (SINR).The relationship between the SINR and the achieved spectral efficiency depends on the technology. The SINR formula is given below: SINR = P (1) I + N where P stands for the power of the incoming signal of interest, I for the interference (baneful signal),and the noise is denoted by N (Figure 1) [6]. Due to the adaptive coding and modulation of the LTE mobile network, different SINR values are assigned to different SE (Figure 2). The envelope of the array of these curves is given by the so-called Alpha-Shannon formula: SE = α log 2 ( IF) SINR (2) where α and IF are constants: α = 0.7, IF = [7]

3 Joint dimensioning of outdoor HetNets using Monte Carlo simulation 137 Figure 1: SINR calculation Figure 2: LTE adaptive coding [8] A. Radio propagation model The SINR formula heavily relies on signal propagation models, calculating both the received useful and interfering signal power. We have applied the widely used, state of the art radio propagation model, the COST231-Okumura- Hata model to calculate the signal rates [9][10][11]. The model can determine the path loss by the help of the frequency, the height of base station, the height of mobile equipment, the distance of between base stations and the type of area. Due to length limitations of this paper, and since the model is well understood in the literature, we will not describe the calculation steps in details. B. Traffic demand: the Monte Carlo Simulation After introducing the way to translate bitrate to radio bandwidth, the next step is to take into consideration the traffic demand for the radio access network. We have used Monte Carlo Simulation to model the users of the network and their traffic demand, creating a series of snapshots of users randomly placed in the appointed area. These snapshots are used to test whether a certain network dimensioning (i.e. antenna densities) are able to meet the traffic requirements [12]. In each snapshot created during the Monte Carlo Simulation, different types of users are placed all over the area with equipartition. During the dimensioning, two types of users are defined. One of them is the simple active user which needs normal traffic demand and the other one is the heavy user which needs more traffic demand than the active user. Then the number of LTE macro base station and LTE pico base station or Wi-Fi AP is analyzed in order to ensure the necessary bandwidth of the users.

4 138 P. Ratkóczy, A. Mitcsenkov C. Antenna density Once a snapshot of users was created during the simulation, with respect to their spatial distribution and bitrate needs, the assignment and connection of the users to the macro base stations or pico base stations/wi-fi APs is examined. First, the system tries to connect all users to a nearby pico base station (Figure 3). If it is not possible because the user is placed far away from the pico base station, or the pico base station has insufficient bandwidth, the user will be connected to the nearest available macro base station (Figure 4). The last step is the evaluation of the given snapshot, in order to determine whether the traffic demand can be served by the given dimensioning parameters (radius of macro base stations, density of pico base stations or Wi-Fi APs). If the given dimensioning parameters cannot serve the input traffic capacity demand, than the number of pico base stations/wi-fi APs have to be increased. Of course, these randomized snapshots are created and evaluated in an iterative process. With certain predefined thresholds, we ensure convergence to coherent results, excluding the effect of randomization. This way the necessary pico cell or Wi-Fi AP densities are determined for various macrocell radiuses, and this way, different HetNet configurations are defined to meet the same traffic demand, with different splits between the macro cell layer and the supplementary technology. Figure 3: Connected to a pico base station Figure 4: Connected to a macro base station

5 Joint dimensioning of outdoor HetNets using Monte Carlo simulation Results In order to get HetNet dimensioning results, we have implemented a MATLAB module based on the models described in the previous sections. In this section we present numerical results, using the reference parameters given in Table 1. Table 1: Input parameters Predefined area: 4x4.5 km (dense urban) Active user density (ρ active ): 240 users/km 2 Heavy user density (ρ heavy ): 60 users/km 2 Active user traffic demand (C active ): 0.5 Mbps Heavy user traffic demand (C heavy ): 2 Mbps A. Joint dimensioning The primary result of the joint dimensioning can be seen on the Figure 5, which is calculated by the Matlab module we have implemented. Namely, we get different combinations of macro and small cells, all if which serve the same traffic demand. Clearly, decreasing the macro cell density requires increased small cell densities but calculating proper combinations is not straightforward. Here, among others, we have considered interference among the cells (not only neighbouring macro cells, but also between macro and small cells); and optimal assignment of users to antennas which altogether gives the opportunity of jointly dimensioning the different layers of the heterogeneous network. Number of pico base stations per macro base staion Mbps/km Number of total LTE macro base stations Figure 5 Result of dimensioning LTE+pico heterogeneous mobile network

6 Number of pico base stations in one macro cell 140 P. Ratkóczy, A. Mitcsenkov B. Equivalent macro and small cell configurations Knowing such equivalent combinations of macro and small cell densities, designed for the same traffic demand, a natural question arises: can we give at least an estimate of the relation between them? Analysing the results we got, we found an inverse logarithmic relation. Such a logarithmic curve fitting is shown on Figure 6. The equation of the logarithmic curve can be written in the following form: P M = α ln(m) + β (3) where P stands for the number of pico base stations, M is the number of macro base stations, α and β are constants. From the above equation, the relationship between the number of pico base stations and the number of macro base stations can be determined: P = α M ln(m) + β (4) Due to the logarithmic behaviour of the curve, the nature of this relation is between the linear (P M) and the square (P M 2 ) functions, i.e. decreasing the macro cell density by a factor of α needs higher small cell densities by a factor of δ (α < δ < α 2 ). 240Mbps/km y = -4,973ln(x) + 24, Number of total LTE macro base stations Figure 6: Curve fitting for macro/small cell densities

7 Joint dimensioning of outdoor HetNets using Monte Carlo simulation 141 C. LTE + Wi-Fi vs. LTE + Pico cell outdoor heterogeneous mobile networks Dimensioning of the LTE macro + Wi-Fi heterogeneous mobile network was carried out with respect to the same capacity parameters. The results are depiced on Figure 7. A significant difference can be recognized between LTE macro + Wi-Fi and LTE macro + pico cell heterogeneous mobile network dimensioning results, with respect to the slope of the curves. As it can be seen, as we start decreasing the density of macro base stations, the amount of necessary Wi-Fi APs rapidly grows, compared to the pico base stations in the same area. The notable difference is primarily caused by the poorer signal propagation characteristics, and hence, the coverage of the Wi-Fi APs than that of the pico base stations. There is a good reason to keep the ISM band free: the radio signal attenuation is a fundamental limitation of the Wi-Fi technology in an outdoor setting. In indoor applications, such disadvantage of Wi-Fi is almost negligible: both Wi-Fi and small cells are able to provide coverage to one indoor space. On the other hand, the building walls that separate an indoor area (building) from another one, blocks the radio signal in any case, both for Wi-Fi and small cells. Therefore, in an indoor setting, the antenna densities are far less sensitive to coverage and signal propagation, which keeps Wi-Fi a competitive option. Number of pico base stations and Wi-Fi APs in one macro cell Mbps/km2 - Pico vs Wi-Fi Wi-Fi Pico cella LTE macro base station density [makro bs/km2] Figure 7: Antenna densities for LTE macro+wi-fi vs. LTE macro+pico HetNets

8 142 P. Ratkóczy, A. Mitcsenkov D. Effect of the input capacity parameters As part of the dimensioning studies, we have also investigated the effect of different traffic patterns on the network dimensioning, i.e. the impact of user density and individual bitrate variations. The curves depicted on Figure 8 show a representative case study: starting from the reference parameters of Table 1, we have doubled the traffic density (Mbps/km 2 ). We can double the traffic in two dimensions: via higher user density, or higher bitrates by each single user. Obviously, the required antenna densities grow, even though somewhat less than twice. The interesting aspect is the comparison of the two dimensions of traffic growth: as it can be seen, individual traffic growth requires more pico base stations than increased user density. This phenomenon can be explained by the batched bin packing problem in the user-antenna assignment phase. When trying to assign users to antennas, having higher user bandwidth demands leads to more unused, fragmented parts of the frequency blocks while smaller, but more densely present users keeps fragmentation lower. Number of pico base stations per macro base staion Changing the input capacity parameters Number of total LTE macro base stations Figure 8: Impact of traffic pattern variations higher density higher capacity

9 4. Conclusion Joint dimensioning of outdoor HetNets using Monte Carlo simulation 143 In this paper, we presented the joint dimensioning aspects of two outdoor heterogeneous mobile networks (HetNets), namely LTE macro cells complemented by either LTE small cells or Wi-Fi technology. We have performed a series of simulations and dimensioning calculations with respect to different traffic parameters. We have investigated the combinations of macro and small cells that are able to handle the same traffic demand, and we have investigated the relation between the density of macro and small cells in the configurations of same traffic capacity: we found an inverse logarithmic relation between macro and small cell densities. We have compared the two presented HetNet solutions (i.e. LTE macro + pico and LTE macro + Wi-Fi) for outdoor use, and we concluded that in outdoor environments, using LTE macro + pico mobile network is suggested, due to physical limitations of signal propagation in the Wi-Fi technology (and frequencies). Finally, we have observed how network dimensioning is affected by changes in the traffic requirements, and we have shown that increasing the number of users has lower impact on antenna densities then higher per user bitrates. Acknowledgement This work has been supported by HSNLab, High Speed Networks Laboratory of the Budapest University of Technology and Economics, The work leading to these results has received funding from the European Union's Seventh Framework Programme (FP7/ ) under grant agreement n COMBO project. References [1] R. Ratasuk, M. A. Uusitalo, N. Mangalvedhe, A. Sorri, S. Iraji, C. Wijting és A. Ghosh, "License-Exempt LTE Deployment in Heterogeneous Network", IEEE ISWCS, [2] M. Bennis, M. Simsek, W. Saad, S. Walid, M. Debbah és A. Czylwik, "When Cellular Meets WiFi in Wireless Small Cell", IEEE Communications Magazine, [3] D. López-Pérez, I. Güvenc, Guillaume de la Roche, M. Kountouris, T. Q.S.Quek és J. Zhang, Enhanced Intercell Intrerference Coordination Challenges in Heterogeneous Networks IEEE Wireless Communications, [4] V. Jungnickel, K. Manolakis, W. Zirwas, B. Panzner, V. Braun, M. Lossow, M. Sternad, R. Apelfröjd és T. Svensson, The Role of Small Cells, Coordinated Multipoint, and Massive MIMO in 5G IEEE Communications Magazine, [5] IEEE Standards Association, a, b/g/n, and ac

10 144 P. Ratkóczy, A. Mitcsenkov [6] D. Chafekar, V. Kumar, M. V.Marathe, S. Parthasarathy és A. Srinivasan, Capacity of Wireless Networks under SINR Interference, Wireless Networks, [7] A. Alexiou, C. Bouras, V. Kokkinos, A. Papazois és G. Tsichritzis, Spectral Efficiency Performance of MBSFN-enabled LTE Networks IEEE WiMob, [8] I Törös, Péter Fazekas, An energy efficient cellular mobile network planning algorithm, IEEE VTC Spring, 2011 [9] Y. Okumura et al., Field strength and its variability in VHF and UHF land-mobile service Rev. Elec. Comm. Lab., pp , [10] M. Hata, Empirical formula for propagation loss in land mobile radio services IEEE Trans. Veh. Tech., pp , [11] E. Damosso, L. M. Correia, Digital Mobile Radio Towards Future Generation Systems Communications, COST231 Final Report, Brussels, Belgium, [12] S. Raychaudhuri, Introduction to Monte Carlo Simulation, in IEEE WSC Simulation Conference, 2008.