Handover Mechanisms for Planned Cell Outage in Twin State Green Wireless Networks

Mythri Hunukumbure, Rajni Agarwal, Sunil Vadgama

Fujitsu Laboratories of Europe Ltd., Hayes Park Central, Hayes End Road, Hayes, Middlesex, U.K. UB4 8FE {Mythri.Hunukumbure, Rajni.Agarwal, Sunil.Vadgama}@uk.fujitsu.com
AbstractThis paper investigates possible handover mechanisms for a twin state network aimed at energy saving. We propose the twin state network as a viable Green option and illustrate the possible energy savings. The challenge of conducting multiple handovers within the transit times of the twin state network, while maintaining the service quality is discussed. The possible handover solutions through multi-carrier operation, FFR and incremental coverage extension are presented in the context of LTE standards. Keywords Green wireless, twin state network, multi carrier handover, FFR handover, incremental handover

handovers involve multiple active users changing their connectivity during the short span of transition periods. This is a type of forced handover and it should be executed in a way that minimizes handover failures. We make use of the availability of multiple carriers in 4G networks (especially the carrier aggregation in LTE [4]) and the facility to employ Fractional Frequency Re-use (FFR) [5] to propose handover mechanisms based on frequency separation. We also look at handover options based on incremental extension of coverage and compare the pros and cons of these approaches. The remainder of this paper is organized as follows. In section II, the twin state network is introduced, followed by a case study which aims to demonstrate the possible energy savings. The frequency separation based handover mechanisms are introduced in section III. This includes both the multicarrier and FFR based approaches. Section IV details the handover mechanisms powered by incremental extension of the coverage. Finally in section V, the conclusions and the further work from this study are drawn up. II. ENERGY SAVINGS FROM TWIN STATE NETWORKS

I.

INTRODUCTION

In view of the stark predictions on climate change and the current CO2 emission levels from the ICT industry, it is widely accepted within the industry that greener technology should be phased in within the next decade. The mobile wireless industry is at the threshold of an exponential increase in service demand due to the advent of mobile broadband applications. This makes it vitally important that greener technologies are developed now, so that the energy per data bit can be contained within sustainable limits in future. The Green radio work item of the Mobile VCE [1] and the GreenTouch project [2] are two of the major initiatives where the industry and academia are collaborating to develop effective greener solutions. When the total energy consumption of a mobile wireless network is considered, the energy consumed at the base stations (both for radio access and backhaul) plays a significant role. Some estimates [3] put the energy consumed by the base station as high as 60% of the total network energy budget. Within this context, it is imperative that the Green wireless solutions address the base station energy consumption issue. We analyse a twin state network, which will operate a higher density of active base stations when the capacity demand is high and lower the active base station density when the capacity demand is low. This involves switching ON and switching OFF some base stations, as governed by the varying capacity demand. We will show that when all the aspects are considered, like the inter-sector interference and backhaul, the energy budget for the twin state network offers considerable savings. The main focus of this paper is the discussion of possible handover mechanisms applicable to a twin state network. The

Radio network planning for cellular networks is typically carried out to meet both coverage and peak capacity requirements. Generally, base-station density is high in populous areas like cities and town centers to meet the heavy demands of mobile voice and data traffic, especially during peak hours. However, the capacity requirement follows both short and long term temporal variations whereby the capacity demand can fall to less than 25% of the peak demand for nearly 8 hours of the day [6]. During such off-peak traffic times, it is potentially possible to achieve energy saving by switching off some base-stations leading to larger cell sizes covered by those that remain on. While the actual saving in the energy is dependent on a number of factors and will vary from case to case, a realistic scenario is illustrated in the subsequent case-study to quantify the gains associated in using a fewer number of base-stations. A. Case study Power consumption model The following case study illustrates power consumption of a cellular network in the twin states. When changing the states, only the transmit power and antenna down-tilt are impacted.

19 micro sites with 43dBm RF power and 4 antenna down-tilt, covering 3 sector cells of 700m radius. This simulates the state with peak traffic.
Energy Saving Ratio (%)

60 50 40 30 20 10 0 24:0 20:4 16:8 12:12 8:16 4:20 0:24

7 of the sites changed to Macro cells which cover the same region with 46dBm RF power and 1.7 antenna down-tilt. The 3 sector cells are now with 1400m radius, operating in the state with off peak traffic.

The power consumption is calculated assuming a typical LTE base-station. As per the linear power model developed in [7], the stand-by (or sleep mode) power rating in the OFF state can be approximated to be 10% of the input power. Also, the input power varies at a much lower rate than the output RF power. In the example scenario, the RF power reduces to 50% in the micro-cell (peak demand) mode, but this will only save about 25% of the input power. Assuming the input rating for a base-station is maximum for the transmit power of 46dBm, the power consumption for 43dBm can be assumed to be approximately 75% of this maximum rating. We have also looked at the micro-wave backhaul power consumption and typically this can be approximated to 4% of the base station input power [8]. Thus the power consumption ratios for the two states defined with the above parameters can be simplistically computed as below: Let PBH be the input power required for the maximum RF output of 46dBm and Plow the input power required to produce 43dBm RF power. Thus we assume; Plow = 0.75*PBH (1) As 19 base-stations are operational in State 1 (peak traffic) and 7 base-stations are operational in State 2 (off-peak traffic), then system power consumptions (BS operational power and backhaul) for two states P1 and P2 are: P1 = 19* 0.75*PBH + 19*0.04*PBH P2 = 7*PBH + 12*0.1*PBH + 7*0.04*PBH (2) (3)

Twin State Operation, Split in hours for 24 hour day (TBH : Tlow) Plow=0.75PBH Plow=0.6PBH Plow=0.95PBH
Figure 1 Energy Savings for Twin State Operation

Plow=0.5PBH Plow=0.8PBH

B. Case study cell capacity evaluation We use the network planning tool Atoll [10] to evaluate the possible network capacity levels achievable with the two states described above. An LTE network is created with 2100MHz operating frequency and 5MHz bandwidth, with the 15 Adaptive Modulation and Coding (AMC) schemes supported in LTE. For this evaluation we only consider the downlink operation. The cell/sector capacity evaluation in state 1 (for the 19 active cells) of the network is illustrated below in Fig. 2.

We have assumed the power required for cooling as negligible. However in certain cases it may be significant. If the duration of operation in State 1 is denoted by TBH and that in State 2 is denoted by Tlow, then based on ETSIs power model [9] power consumption for system can be denoted by: Ptwin = (P1 * TBH + P2 * Tlow)/(TBH + Tlow) (4) Thus, power saving in a twin state network can be quantified as: PStwin = [(P1 Ptwin)/P1]*100 (5) The power saving achieved by the twin state network will vary as per the time period the network is able to operate in state 2, Tlow. Considering different proportions of TBH:Tlow for the 24 hour duration of a day, the power saving achieved in a system operating in twin states as against the traditional single state 1 is illustrated in the Fig. 1 below. As the Plow value (as a proportion of PBH) for different base station models can vary, we also include the possible energy savings for different ratios of (Plow/PBH) in Fig. 1. For the adopted ratio of Plow/PBH =0.75 in the case study, the twin state network offers around 15% power saving, considering 8 hours operation (per day) in the off-peak traffic state.

Figure 2. Cell capacity evaluation for the state 1 of the LTE network

The capacity in each coloured region in the map is related to the SINR in that region and the AMC scheme it can support. The interference power plays a crucial role here. This is demonstrated by the fact that the inner cells/sectors are conditioned to regular shapes while the sectors at the network edges expand widely due to lack of inter-sector interference. A typical network will contain many more cells, so we only use the capacity of the inner sectors to evaluate the network capacity. The sector capacity is calculated by considering the percentage coverage of each coloured region of a typical inner sector. This comes upto 6.8Mbps in this illustration. Hence the total network capacity for the 57 sectors can be averaged at 387.6Mbps. The cell/sector capacity evaluation for the off-peak state 2 of the example network is illustrated below in Fig. 3.

III.

HANDOVERS BASED ON FREQUENCY SEPARATION

Within the transition period of a twin state network, the connections of active users need to be switched from one BS to another, without handover failure. Both LTE (release 10) and the WiMAX .16m standards support multi-carrier operations and the use of FFR. Hence the use of multiple, forced handovers based on frequency separation in a twin state network are well within the capabilities of future 4G systems. A. HO in a multi-carrier setting Multi-carrier operations will become a common feature of 4G networks, as this enables the networks to operate in increasingly fragmented spectrum allocations. For example, carrier aggregation (CA) is an important theme discussed in LTE standardisation [4]. In an LTE set-up, there will be primary and secondary carriers available to the BS (or eNodeBs in LTE terminology). The primary carriers operate with the full signalling load (full control channel occupancy) while the secondary carriers have a reduced signalling load. In a twin state network, the mechanism depicted in Fig. 4 can be used to switch the active users from a cell which is to be disabled to another cell extending its coverage. It is assumed that in two adjacent cells served by BS1 and BS2 both base stations employ a primary carrier at frequency f1 and a secondary carrier at frequency f2. The height of the boxes indicate the transmit power while the width indicates the allocated carrier bandwidth.

Figure 3. Cell capacity evaluation for the state 2 of the LTE network

The coverage regions for each sector has considerably expanded in state 2, due to the higher transmit power (46dBm) and the lower antenna down-tilt. In terms of cell capacity the main difference is that at cell/sector edges, the network can only support 10kbps services. This is equivalent to basic voice quality in LTE. We believe that the requirement for high QoS at these cell edges would be significantly low at off peak periods. In our view, this is a fair price to pay for energy savings that can be achieved. The sector capacity is calculated for an inner sector as before. The typical sector capacity is now 5.6Mbps, which is 17.5% lower than the sector capacity for state 1 of the network. The total network capacity with the 21 active sectors is 117.6Mbps. Comparing the two states, the off-peak capacity supported by the network is roughly 30% of the peak capacity. Hence our illustrated twin state network can be operated when the network load drops below 30% of the peak demand.

Figure 4. Proposed handover procedure for a multi-carrier system

The handover procedure can be executed in 3 steps as follows; In both the adjacent cells served by BS1 and BS2, users are transferred from the secondary carrier to the primary carrier. This can be a formal inter-frequency handover or a scheduling procedure, depending on the network set-up. Once the transfer is complete, both BS1 and BS2 switch off their secondary carriers. BS1 configures a new primary carrier B on the vacant f2 band, providing extended coverage to cell 2. BS2 hands over all its active UEs to the new primary carrier in inter-frequency handovers. BS2 is disabled. All its active users are now served by primary carrier B, with extended coverage from BS1.

Within the LTE context, the inter-frequency handovers of users from BS2 to BS1 are based on measurement reports provided by the user. The specific measurements can be triggered by event A3 [11] where the neighbour cell operating on a different frequency plus an offset becomes stronger than the serving cell. The entering condition and leaving condition for the A3 measurement is shown below (from [11]). Entering condition;
Mn + Ofn + Ocn Hys > Ms + Ofs + Ocs + Off

primary carrier A, and active users connected to BS2 will be prompted to handover to this primary carrier A, by controlling the offset parameter Off. Once all the handovers are complete, the temporary carrier can be used elsewhere. B. HO in a FFR setting Fractional Frequency Re-use (FFR) offers a tradeoff between the high spectral efficiency of a frequency re-use 1 option and better QoS for the cell edge users with some reduction in interference. The current trend in FFR deployment for both LTE-A and WiMAX .16m is soft FFR, where all the available spectrum is used by every cell, but certain selected bands are used with a much lower power so the interference they generate is minimal [5]. We propose a handover procedure applicable in a soft FFR setting. The procedure is illustrated in Fig. 5, executed in 4 phases. As in Fig. 4, the heights of the boxes indicate the transmit power while the width indicates the bandwidth of the frequency partitions used in soft FFR. The bandwidths of the partitions in practice will mostly be different (tailored to suit the requirements) but in order to maintain clarity we assume identical bandwidths.

(6) (7)

Leaving condition;
Mn + Ofn + Ocn + Hys < Ms + Ofs + Ocs + Off

Where Mn and Ms refer to neighbour cell and serving cell signal strength or signal quality measurements. Ofn and Ofs are frequency specific offsets for the neighbour and serving cells. Ocn and Ocs are cell specific offsets for the neighbour and serving cells, while Hys is an hysteresis value to prevent the ping-pong style handovers. Off denotes the basic offset between the serving cell and neighbour cell measurements. At cell edge (of BS1 and BS2) the coverage from the serving BS2 and the neighbour BS1 (to be extended) are roughly equal, so by setting the offset Off closer to zero the above A3 event can be triggered and the measurements reported to BS2. BS2 can then initiate the handover to BS1 for the users near this cell edge. As one moves inside the cell of BS2, its own signal strength gets stronger than the extended coverage from BS1. Hence the offset needs to be set to decrementing negative values for the event A3 and subsequent handover to occur. Thus by controlling the value of Off, active users in different regions of cell BS2 can be handed over. This prevents an avalanche of handover conditions being created for all the active users in BS2 at the same time. The LTE standards allow the parameter Off to be varied from -30dB to 30dB [11]. In LTE, the mobility control of the idle users is user centric. The users maintain knowledge of their tracking areas in idle mode, where the tracking area consists of multiple cells. In order to avoid over-the-air signaling from a large number of idle users, we propose the cell to be extended and the cell to be disabled should belong to the same tracking area. The main advantage of this type of inter-frequency handovers is that the interference effects are minimized, due to the separation of the two carriers involved. Also, the time taken to handover all the users can be controlled by the speed of adjusting the offset Off values. Another advantage with the multi-carrier approach is that it allows the antenna down-tilt and the transmit power to be adjusted optimally per carrier, to best serve the cell (or segment of cell) concerned. The need to have a secondary carrier is a constraint in this handover solution, although many LTE networks will have secondary carriers due to carrier aggregation policy [4]. Even though permanent secondary carriers are not available, a similar solution can be implemented by using a temporary carrier. In this case, the BS1 and BS2 (in Fig. 4) will utilize only the primary carriers A and C. BS2 will launch a temporary carrier on f2, where all the active users in BS2 will be transferred to f2. Then BS1 will extend the coverage of its

Figure 5. Proposed handover procedure for a soft FFR system

Both BS1 and BS2 have the same frequency reuse 1 (FR1) partition P0, but alternate FR3 partitions P1 and P2 are powered up or down. These 4 steps describe the handover procedure. Users are transferred from the FR3 partition P1 in both BS1 and BS2 to the FR1 partition of P0. The transmit power of P0 partition in BS1 is increased to cover the cell edge for this purpose. This transfer is a scheduling procedure. Once the user transfer is complete the partition P1 will become empty in both BS1 and BS2. The partition P1 is extended from BS1 to cell 2 with increased transmit power. The antenna tilt may also be adjusted (of all the partitions as they are bound to a single carrier) to optimize this extended coverage. The active users in the P0 and P2 partitions of BS2 are handed over to the P1 partition of BS1. BS2 is disabled.

As in the case of inter-frequency handover, this FFR based handover can be controlled by varying the Off value in user measurements in an LTE setup. The trigger for measurements will be controlled as per equations (6) and (7) but this time the values for Ofn and Ofs will be identically set as these are intrafrequency handovers. By gradually changing the value of Off from near zero to decrementing negative values, active users from cell-edge to inner cell of BS2 can be systematically handed over. This solution minimizes any interference effects at handover, as the available frequency partitions are separated for the active users in BS1 (cell1) and the disabled BS2. The handover speed (for the whole operation) can be controlled by the speed of the value decrement for parameter Off. One of the draw backs of this HO procedure is that it is difficult to optimize coverage for both the serving cell 1 and the BS disabled cell 2 with a single carrier. The transmit power and antenna down-tilt adjustments need to provide a compromise coverage for both cells. IV. HANDOVER WITH INCREMENTING COVERAGE

because of the vast difference in the path loss between the two signals. In this case, the handover has to occur with a negative offset Off, as the signal levels in Fig. 6 suggest. Thus after the coverage of cell 1 is fully incremented and of cell 2 fully decremented, the offset values of the handover measurement process need to be set to negative values, so the remaining UEs can complete their handovers. The main advantage of this handover scheme is the simplicity of operation. It does not require separate carriers or FFR, or the additional user transfers associated with them. The downside is the possible interference effects and the resulting handover failures, when using the same resource for mass handover. V. CONCLUSIONS AND FURTHER WORK

In certain networks, there may arise situations where neither spare carriers are available nor FFR is enabled. In such cases, a simpler form of forced handover can be implemented. This will increment the coverage of the extending cell while decrementing the coverage of the cell to be disabled. This coverage adjustment can be carried out in steps, so that handover zones are created inside the cell to be disabled. This process is illustrated in Fig. 6 below.

In this paper we discuss the handover mechanisms in twin state networks, which change the active site density according to the network load. We put forward the case for twin state networks to be a viable option as a Green wireless solution. Energy consumptions are compared in an example twin state network, showing the possible energy savings. The operating modes for the proposed handover mechanisms are discussed in a LTE context with relative pros and cons. Looking forward, the next step to consider is to examine the relative performance of the handover mechanisms in a LTE System Level Simulator (SLS), so that a quantitative assessment can be made. We will also consider the variations of eNB power consumption with varying load conditions to enhance the power consumption model. REFERENCES
Mobile VCE core 5 Green radio research theme, http://www.mobilevce.com/frames.htm?core5research.htm [2] Green Touch consortium, http://www.greentouch.org/ [3] Vodafone sustainability report FY 2009-10, available at http://www.vodafone.com/etc/medialib/cr10/pdf.Par.17290.File.dat/voda fone_sustainability_report.pdf [4] LTE-Advanced Carrier Aggregation of Band 3 and Band 7, work item documents available at http://www.3gpp.org/ftp/specs/htmlinfo/FeatureOrStudyItemFile-480023.htm [5] N. Himayat et. al., Interference Management for 4G Cellular Standards, IEEE Comms Magazine, Vol. 48, No. 8, August 2010, pp. 86-92 [6] Sustainable energy use in mobile communications, Ericsson White paper, August 2007, available at http://www.ericsson.com/campaign/sustainable_mobile_communication s/downloads/sustainable_energy.pdf [7] L. Correia et. al., Challenges and Enabling Technologies for Energy Aware Mobile Raion Networks, IEEE Comms Magazine, Vol. 48, No. 11, November 2010, pp. 66-72 [8] D. Coudert et. al, Wireless Backhaul Networks: Minimizing energy consumption by power efficient radio links configuration, INRIA report, available at http://hal.archivesouvertes.fr/docs/00/36/39/21/PDF/RR-6752.pdf [9] ETSI TS 102 706 v1.1.1, Environmental Engineering (EE) Energy Efficiency of Wireless Access Network Equipment, August 2009. [10] Atoll product details available at http://www.forsk.com [11] 3GPP LTE specification 36.331 v9.2.0, Radio Resource Control; Protocol Specification, section 5.5.4.4. pp75-76. [1]

Figure 6. Handover process with incremental changes in coverage

As per the diagram, 4 users A, B, C and D in the cell 2 (cell to be disabled) need be handed over to the cell 1 (cell with extended coverage). The handover procedure is explained with the aid of a signal strength graph in Fig. 5. As the coverage of cell 1 is incremented and cell 2s coverage is decremented, it creates handover points as shown. Initially the offset value Off (equations (6) and (7)) can be set to a positive value, so the handover occurs only when cell 1s incrementing coverage becomes better than cell 2s coverage. While this may lead to some UEs to radio link failure due to increased interference, it generally allows a stable handover without ping-pong effects. Users B, C and D may be handed over in this fashion. However for user A, who is nearest to the cell centre of cell 2, the incrementing signal level of cell 1 may not match even the lowest decremented level of cell 2. This is