A Dual-hop Backhaul Network Architecture for 5G Ultra- Small Cells Using Millimetre-Wave

Transcription

1 A Dual-hop Backhaul Network Architecture for 5G Ultra- Small Cells Using Millimetre-Wave Aftab Ahmed and David Grace Communications Research Group, University of York, York, YO0 5DD, United Kingdom Abstract: This paper instigates a new network architecture for future 5G backhaul networks which can be used to achieve higher throughput capacity and better Quality of Service in high data traffic scenarios. This network is predominantly based on dual-hop wireless links which introduce an intermediate relay point between backhaul transceivers. To make this network a high capacity density wireless system, the base stations are densely deployed and equipped with highly directional antennas which use millimetre wave. It is shown that the overall performance can be improved by up to 5% by reducing the bottleneck and introducing routing diversity to such a dense dual-hop backhaul network. Moreover, it is shown how the performance of the dual-hop backhaul network can be modelled using the Erlang-B traffic distribution. Keywords: Dual-hop, backhaul networks, millimetre wave, routing diversity, ultra-small cells. W I. INTRODUCTION ith the escalation of next-generation electronic gadgets (mobile phones, tablets) and with the commencement of data intensive applications, it is forecast that global data traffic is expected to proliferate to. Exabyte per month by 09 [] mainly because of video streaming and use of cloud computing on smartphones and tablets. The tenfold increase over 0 is anticipated to continue at least until 00, during the life of 5 th -Generation (5G) networks. For this reason next generation wireless networks will be required to provide data access rates up to 0 Gbps and have a round trip delay of less than millisecond (ms), with users having a high degree of mobility []. Transferring such high levels of delay sensitive data from the access network to an aggregation point (equipped with fibre) via one or more hops is considered to be a very hot research topic and is examined in this research work. One way of delivering such high data rate requirements, is with very dense deployments of ultra-small cell Access Base Stations (ABSs). Data traffic is transferred from these ABSs to the Hub Base Stations (HBSs) via dualhop backhaul links through the introduction of intermediate Relay-Access Base Stations (R-ABSs) between ABS and HBS as shown in figure.. The number of hops for any connection is restricted to two. Any end-to-end connection is established on link-bylink basis. A single link is defined as a connection between any two transceiver points []. Due to this dense deployment, backhaul and routing requirements, it is predicated that the 5G networks will be very complex and will need to be equipped with self-healing and selforganizing capabilities, thus reducing the need for manpower. In these 5G ultra-small cell networks the aim is that the end user has the perception that they are connected to the core network []. Figure.. A Dual-hop Backhaul Network Model Point-to-point microwave links are already used with G networks, which makes the densification of BSs proposed for 5G even more substantial. This densification then leads to many ultra-small size cells for communication in an urban environment. In this paper, point-to-point communication links between the base stations (BSs) are restricted to those using the millimetre wave (mmwave) E- band which have the opportunity to exploit large contiguous bandwidths in the frequency ranges 7-76 GHz and 8-86 GHz [5] that can be used for outdoor urban backhaul communications [6]. Potentials and applications of millimetre wave are already demonstrated in [7] and [8] for 5G cellular networks. The capabilities and benefits of dualhop networks to achieve higher Quality of Service (QoS) are also shown in [9]. In future 5G ultra-small cell networks, the number of hops on the backhaul side can be more than one /5/$ IEEE

2 for achieving better QoS [0]. Extensive research campaigns are carried out in [6] and [8] on various aspects of mmwave to establish a better insight and to confirm its validity for the future 5G communication networks. The remaining paper is organized as follows: Section II describes the novel dual-hop network architecture. Section III explains the Routing Process. Section IV discusses the Performance Analysis. Section V demonstrates Simulation Results. At the end, conclusions are drawn in section VI. II. PROBLEM DESCRIPTION A. Network Architecture The network architecture which is proposed in this paper is an extended form of Manhattan grid like structure but with a dense deployment of ABSs as shown in figure.. It can be considered as one of the contenders for a future 5G backhaul network architecture. In this model, the R-ABS is introduced to relay the data traffic from ABS to the HBS with the number of hops to be equal to two anywhere in the network. In this example, the network service area comprises of 50*50 meters (m) as shown in figure.. This network consists of 6 ABSs which are installed outdoors along the street at each street crossing (e.g. street lamps, traffic lights) separated by a distance of 90 m. The total number of R- ABSs is also 6 with only one of them installed on each side of a street end. Also, a total of HBSs are deployed with only one on each corner of the block. These HBSs act as aggregation points and are connected to the Operator s Core network via fibre as shown in figure.. The distinctive trademark of this network architecture which makes it to stand out over the traditional backhaul network is its two hop backhaul feature. This dual-hop feature ensures that the latency requirements of future 5G networks can be met, while providing increased flexibility. These dual-hop connections are provided by highly directional antennas mounted on each ABS, R-ABS and HBS transceiver. The Half Power Beamwidth (HPBW) of these antennas is 0 [] which significantly improves the link budget performance, while also mitigating the interference, thereby enhancing the throughput capacity []. Each ABS is capable of generating up to a maximum of four beams for transmission with main-lobe separation of 90 between two beams. At each R-ABS there is only one beam pointing towards four in line ABSs for reception and two more beams for transmission with main-lobe separation of 80 pointing towards the two HBSs deployed on the same street. However, each HBS is equipped with two beams with mainlobe separation of 90. Each beam is capable of reception from the corresponding four R-ABSs deployed on the same street. The link between ABS to R- ABS is the first hop of the dual-hop connection as shown in figure. while the link between R-ABS to HBS is the second hop. To compute the path loss of the transmitted signal as a function of the distance, the close-in reference distance model [6] is used, which is compatible with the dual-hop stationary BS Line-of-Sight (LOS) relay setup. The ABSs and R-ABSs are posted below rooftop height whereas the HBSs are deployed above the rooftops. Further assumptions in this work are given below. The file arrival rate is modelled as Poisson distribution with a fixed mean arrival rate of lambda (λ). GPP FTP Traffic Model [] is used with an fixed file size of Mbytes. In order to see the impact of routing on these dual-hop backhaul links, it is assumed that R-ABS is only capable of relaying data traffic from the ABS towards HBS. The impact of data traffic generated by the R- ABSs will be investigated in the future work. Only the uplink transmissions of the backhaul from the ABSs towards HBSs via R-ABS are considered with a fixed transmission rate and there is no interruption in the network. A perfect access network is assumed which means that blocking and delay are caused solely by the backhaul network []. It is assumed that delay caused by queuing and propagation is negligible compared to the transmission time. (m) Step # Step # (m) HBS R-ABS ABS Buildings Figure.. Network Architecture The zoomed region in figure. illustrates the two steps of Routing Process explained in section III. West North South East

3 B. Resource Assignment Scheme The frequency bandwidth of 500 MHz is divided into 00 equally sized (.5 MHz) channels. With the deployment of highly directional antennas using mm-wave and having short distance links, the link budget is significantly improved, thus allowing the same frequency bandwidth to be used on both the links. Dynamic Channel Allocation strategy is used for resource assignment while frequency channels are assigned in sequence on First Available basis at both links separately. The resource assignment scheme is centralised where two separate channel state tables are maintained, one for all R-ABSs and other for all HBSs. A file is only initiated by the source ABS and transmitted to the destination HBS via R-ABS. A Dual-hop end-to-end connection can only be established if there is at least one free channel per hop [], otherwise the transmission is blocked as shown in the flow diagram in figure.. There is end-to-end signalling mechanism on the backhaul network which makes sure that the Signal Interference plus Noise Ratio (SINR) is above the required threshold in order to establish the dual-hop connection. In case of blocking, the retransmission only starts after a random time interval, provided that a free good quality channel is available in the system. Any link can be assigned multiple channels, once there is more than one file transmission simultaneously. nd hop channel available? Yes No Yes st hop channel available? No Transmission blocked No File arrival Retransmission later Transmission started ABS selects the next closest R-ABS on the North-South Street (marked ) and so on as illustrated in figure.. st closest R-ABS: West nd closest R-ABS: North ABS R-ABS st closest R-ABS: West nd closest R-ABS: South Figure.. Routing Process ( st Step) st closest R-ABS: East nd closest R-ABS: North st closest R-ABS: East nd closest R-ABS: South nd Step: This step is associated with the second link of the dual-hop connection. The R-ABS also selects a free channel from any one of the two HBSs deployed on the same street where the transmitting R-ABS is installed. This time the R- ABS selects a free channel from the closest HBS (marked ) as shown in figure 5. In case if no free channel is available from the closest one, the R-ABS moves to the second HBS on the opposite end of the street (marked ) as shown in figure.5. Once a free channel is available on both hops of the connection, the end-to-end connection is established and the data transmission gets under way. It is noteworthy that a bottleneck occurs on the second link of the dual-hop connection; therefore the routing process ameliorates the performance mainly on the second hop by introducing more routes to the aggregation point HBS. These HBSs are connected to Operator s Core Network via wired backhaul as shown in figure.. The bottleneck effect becomes more pronounced when the R- ABSs forward their own cell data traffic on the second link which is beyond the scope of this paper. Closest HBS: North Closest HBS: South Both channel SINR >= threshold? Yes Dual-hop connection established HBS R-ABS Figure.. Flow Diagram for Resource Allocation (hop by hop) III. ROUTING PROCESS The Routing Process for this dual-hop backhaul network is shown in figure. (zoomed region). Its implementation is carried out in the following two steps. st Step: This step deals with the first link of the dual-hop connection. When a new event arrival occurs, the transmitting ABS selects the closest R-ABS deployed on the East-West Street using any one of its four transmitting beams. Each beam is oriented in any one of the four directions (North, South, East, and West). If there is no free channel available with closest R-ABS (marked ) then the Closest HBS: East Figure.5. Routing Process ( nd Step) Closest HBS: West IV. PERFORMANCE ANALYSIS A Dual-hop Link Model which is shown in figure. is compared with the Erlang-B Model for analytical verification. The GPP file traffic model is designed as such that a single ABS can have any number of simultaneous file transmissions which can be approximated to infinite user population. The bottleneck which occurs on the second hop only is analytically verified using the Erlang-B Model. As shown in

4 figure.6, there is a perfect match between simulated model and the Erlang-B Model. Also, the System Throughput graph becomes limited as the offered traffic increases beyond 50 Mbps due to the bottleneck on the second hop. Also, the Erlang-B Model works on the principle of Blocked Calls Cleared, that s why no retransmission or queuing is assumed for this analytical comparison [5]. Blocking Probability Pb Blocking Probability & System Throughput System blocking Erlang-B blocking System Throughput Offered Traffic (Mbps) Figure.6. Comparison between Simulated Model & Erlang-B Model V. SIMULATION RESULTS The Dual-Hop Network Architecture suggested in section II (A) is simulated using the Routing Process explained in section III. The Routing Process is performed for each number of iteration to establish an end-to-end connection. Other parameters used in this simulation are summarized in Table.. The performance of this network is evaluated using discrete time event based simulation. Files arrive into the system with exponentially distributed interarrival times and leave the system once the file transmission is completed. The simulation environment consists of 6 ABSs, 6 R-ABSs and HBSs in total. All suitable combinations of Routing Process are implemented using the same First Available Resource Assignment Scheme as described in section II (B). Table.. Simulation Parameters Parameter Value Building block size 75 m * 75 m Street width 5 m Max antenna gain 7 dbi Carrier frequency 75 GHz Frequency Bandwidth 500 MHz Channels 00 Single file transmission rate.5 Mbps Transmission power 7 dbm Iterations per offered traffic level Throughput (Mbps) The QoS parameters of this proposed backhaul network are evaluated in terms of Blocking Probability (Pb) and endto-end Mean Transmission Delay. The interruption of the transmission link due to worsening link quality in terms of SINR is not assumed in this work given the highly directional nature of the antennas and mm-wave frequency band which results in minimal interference. The blocking occurs in this network on either link only due to the unavailability of a free channel. The adequate limit for the blocking probability is assumed to be 5%. All the performance parameters are plotted against six different routing combinations. For this simulation the offered traffic, specifically the user demanded traffic is divided into three levels. This dual- hop routing approach could result in many possible combinations. Here this paper examines the impact when the number of possible routes is restricted. The routing choices available are specified using two numbers, the first integer indicates the maximum number of routes that a transmitting ABS has for the first link towards the intermediate R-ABS while the second integer is associated with the number of routes that R-ABS has for the second link towards the destination HBS. The route options are limited on the basis of the shortest distance to all four possible HBSs. All the six combinations of routing are simulated for three different ranges of offered traffic and their corresponding blocking probability values are shown in figure.7. From figure.7 it is evident that it is the second link where the greater routing choice to the HBS improves the performance. It also validates the assumption that the bottleneck effect is taking place on the second link at higher levels of offered traffic, which increases the blocking. There is still slight improvement in the performance if the number of routes on the first link is increased. For the baseline scenario, when there is no routing and path diversity (-) available on both the links, the network is only capable of providing the same QoS up to an offered traffic of 5 Gbps. It is this routing and path diversity which reduces the bottleneck on the second link and improves the QoS by 5%. It is evident that for routing level - and -, the Pb value increases from % to 7% at an offered traffic of 7. Gbps. Figure.8 on the other hand, illustrates the impact of this dual-hop network along with the routing process in terms of Mean Transmission Delay. In case of no blocking in the network, the Mean Transmission Delay is equal to the Transmission Time. Once again, at higher offered traffic the transmission delay significantly increases when there is only one routing option on the second link. Figure.8 also validates the fact that the bottlecneck is occuring on the second hop as shown by blocking probability in figure.7.

5 Blocking Probability (Pb) Blocking Probability vs Routing Level Offered Traffic 6. Gbps Offered Traffic 6.8 Gbps Offered Traffic 7. Gbps the channel index increases, CU % decreases which is in compliance with the Resource Assignment Scheme explained in section II (C). It also confirms that the routing process is independent of Resource Assignment. Those channels which exceed 8 are used only for high offered traffic levels which give rise to the curve shape in the middle of the graph. 0.8 Channel Usage Percentage Routing Level Figure.7. Routing impact on Blocking Probability Transmission Delay vs Routing Level Offered Traffic 6. Gbps Offered Traffic 6.8 Gbps Offered Traffic 7. Gbps Channel Usage Percentage % Mean Transmission Delay (sec) Routing Level Figure.8. Routing Impact on Transmission Delay Figure.9 shows the Perceived Throughput per channel plotted against routing level. The Perceived Throughput decreases by 0% between routing level - and - for an offered traffic of 7. Gbps. These three performance evaluation parameters indicate that by having more routing choices on the second link especially, the performance of this dual-hop network can be enhanced by reducing the bottleneck. It must be noted that the baseline routing level - is not shown in these simulation results because it is incapable of handling anything above 5 Gbps. Perceived throughput (Mbps/CH) Perceived Throughput vs Routing Level Routing Level Offered Traffic 6. Gbps Offered Traffic 6.8 Gbps Offered Traffic 7. Gbps Figure.9. Routing impact on Perceived Throughput Figure.0 shows Channel Usage Percentage (CU % ). It illustrates that the starting channels are used the most but as Channels Figure.0. Channel Usage Percentage Simulation results from figure.7 to figure.9 clearly demonstrate the level of expected improvement in QoS that can be achieved by introducing more routing and path alternatives into the dual-hop backhaul network. Much higher system throughput is achieved by having such close deployment of base stations. Also installing highly directional antennas using mm-wave band for such short distance links, the level of interference can be significantly mitigated. VI. CONCLUSION This paper proposes dual-hop backhaul network architecture for ultra-small cells with path and routing diversity. This provides significant flexibility in limiting the number of aggregation points where fibre needs to be present while ensuring that latency constraints can still be met. With this dual-hop approach it is shown that bottlenecks are restricted to the second hop links but routing flexibility mitigates this bottleneck impact to a larger extent. Furthermore, it is also analytically verified that with this type of dual-hop architecture, the hop closest to the aggregation point is experiencing the bottleneck. It is this unique dual-hop feature of the backhaul network which makes it distinctive from a single-hop backhaul networks. The level of performance is significantly improved by 5% at the expense of routing and path diversity available in such dense dual-hop network. Due to this path diversity, some of backhaul base stations can be turned off at low offered traffic while still maintaining the same QoS thus minimizing the energy expenditure of the whole network.

6 REFERENCES [] "Cisco Visual Networking Index Global Mobile data traffic forcast update, 0-09.," CiscoFebruary 05. [] Nokia, "White paper for 5G requirement.," 05. [] Z. Qiyang and D. Grace, "Application of cognition based resource allocation strategies on a multi-hop backhaul network," in Communication Systems (ICCS), 0 IEEE International Conference on, 0, pp. -7. [] R. E. Hattachi and J. Erfanian, " 5G White Paper," Next Generation Mobile Networks. 05. [5] J. L. G. Cedric Dehos, Antonio De Domenico, Dimitri Kténas, and Laurent Dussopt, "Millimeter-Wave Access and Backhauling: The Solution to the Exponential Data Traffic Increase in 5G Mobile Communications Systems," IEEE Communications Magazine, 0. [6] G. R. MacCartney and T. S. Rappaport, "7 GHz millimeter wave propagation measurements for outdoor urban mobile and backhaul communications in New York City," in Communications (ICC), 0 IEEE International Conference on, 0, pp [7] K. Joongheon and A. F. Molisch, "Quality-aware millimeterwave device-to-device multi-hop routing for 5G cellular networks," in Communications (ICC), 0 IEEE International Conference on, 0, pp [8] S. Rangan, T. S. Rappaport, and E. Erkip, "Millimeter-Wave Cellular Wireless Networks: Potentials and Challenges," Proceedings of the IEEE, vol. 0, pp , 0. [9] W. Hung-Yu and R. D. Gitlin, "WWAN/WLAN two-hop-relay architecture for capacity enhancement," in Wireless Communications and Networking Conference, 00. WCNC. 00 IEEE, 00, pp. 5-0 Vol.. [0] G. Shen, J. Liu, D. Wang, J. Wang, and S. Jin, "Multi-hop relay for next-generation wireless access networks," Bell Labs Technical Journal, vol., pp. 75-9, 0. [] T. Jiang, A. Papadogiannis, D. Grace, and A. Burr, "BuNGee Deliverable..-Interim Simulation," 0. [] A. S. Tanenbaum, Computer Networks, th ed. vol., 00. [] "GPP, "Technical Specification Group Radio Access Network; Evolved Universal Terrestrial Radio Access (E-UTRA); Further advancements for E-UTRA physical layer aspects", GPP TR 6.8 V9.0.0 March 00," ed, 00. [] L. Jialu and D. Grace, "Cognitive green backhaul deployments for future 5G networks," in Cognitive Cellular Systems (CCS), 0 st International Workshop on, 0, pp. -5. [5] J. Yu and I. A. Ajarmeh, "Design and Traffic Engineering of VoIP for Enterprise and Carrier Networks," presented at the International Journal On Advances in Telecommunications, 008