Journal of Communications and Information Networks

Femtocells are considered as a favorable solution for theemerging high-bandwidth requirements of multimedia applications^[1,2]. Together with advanced technologies such as long-term evolution(LTE), it is expected that there will be an increase in the overall network capacity. Femtocells are small, low-power base stations that can be overlaid on existing cellular networks^[3]. To this end, the 3rd Generation Partner ship Project (3GPP) has initiated the deployment of heterogeneous networks in LTE-Advanced specifications. Heterogeneous networks consist of macrocells and low-power nodes such as microcells, picocells, and femtocells^[4]. Femtocells are mainly developed to provide broadband connectivity in indoor environments. The traffic generated from indoor environments suffers from poor signal quality and higher penetration losses, which leads to a requirement for more resources to guarantee the required quality of service(QoS). Femtocells have therefore been introduced to overcome the indoor coverage problem and enhance the cell capacity by offloading the traffic from macrocell Evolved Node Bs(eNodeBs). From an operator’s point of view, this traffic offloading will lead to the need for fewer macrocell sites, which results in huge Capital Expenditure(CAPEX)savings and uninterrupted user experience in both indoor and outdoor environments^[5].

Femtocells operate in the licensed frequency bands, and can be easily installed by consumers upon enabling smallcoverage radio-based stations, also called Home eNodeBs (HeNodeBs). However, the extensive installation of femtocells can cause severe interference-related issues^[6]. Because macrocells and femtocells use shared spectrum, strong crosstier interference can exist between both the tiers. The severity of the cross-tier interference depends significantly on the access mechanism used in the femtocell network^[7]. In access control mechanisms, users can be classified as subscribers who are registered users or nonsubscribers who are random mobile users, depending on the connectivity privileges.

Three access methods have been proposed for femtocells in the 3GPP specifications: closed access, open access, and hybrid access, as shown in Fig.1.

In closed-access mode, only subscribers are granted access to the corresponding femtocells. Although this access method increases the QoS of registered user equipment (UE), it will result in severe interference-related issues. Many studies have been performed to achieve interference mitigation, including power control^[8,9,10,11] , spectrumsharing techniques^[12,13,14], and cognitive approaches^[15,16]. In Ref. [17], the resource-allocation problem in a two-tier orthogonal frequency-division multiple access(OFDMA)based network is studied, where the femtocells employ closedaccess mode and have cognitive radio capability to sense the channel environment. The cognitive radio capability helps the femtocells to choose the subchannels that have minimal interference and that maximize the sum throughput of femtocell users. In an open-access mode, all the users are permitted to connect via the femtocells. Earlier studies^[18,19] have shown that the installation of open-access femtocells improves the system performance by offloading the macrocell traffic. In Ref. [20], a resource-allocation method for openaccess OFDMA femtocells is proposed by using cognitiveradio technology in femtocells. This access mode is used as an alternative approach for interference reduction because UEs will try to connect to the server with the strongest signal. However, this access mode will lead to an increase in the number of handovers, negatively impacting operators.

As a compromise between these two access methods, the hybrid access method is introduced, allowing preferential access for subscribers and limited access for nonsubscribers^[21]. The resource sharing between subscribers and nonsubscribers in a femtocell network needs to be regulated as it is necessary to guarantee the minimum throughput required by subscribers. The remaining resources can then be allocated among nonsubscribers in an efficient way by effectively using different scheduling methods. Nonsubscribers can be served on a first-come first-serve basis, or priority can be given based on their service requirements^[22]. This study investigates resource scheduling for femtocell networks in hybrid-access mode, which gives preferential access to high-priority traffic, thereby increasing the overall average throughput of highpriority users in a network.

The rest of the paper is organized as follows. The system model of hybrid-access femtocell networks is described in section Ⅱ. In section Ⅲ, the proposed resource-scheduling method is explained in detail. In section Ⅳ, the simulation results are discussed. The conclusions are presented in section V.

In this paper, a downlink(DL)OFDMA-based scenario is considered that consists of a single macrocell and between 130 femtocells. The macrocell eNodeB is located at the center, while the femtocells f ∈{1, 2, …, F}are randomly distributed within the coverage area of the macrocell. It is assumed that the total power is equally divided among all the subcarriers. It is also assumed that none of the femtocells overlap with one another or with the macrocell eNodeB. The cochannel deployment is considered in the system, which indicates that both the macrocell and femtocells use shared spectrum. The macrocell eNodeB and all femtocell HeNodeBs are linked to a local entity that controls the resource management centrally in the system, as shown in Fig.2. The local entity obtains the overall channel-state information at the beginning of every transmit time interval(TTI)to determine which base station has to provide service to the UE. This information is reported from base stations to a local entity using a feedback method. It is assumed that all the femtocells f ∈{1, 2, …, F}in the macrocell coverage area operate in the hybrid-access mode, with every femtocell serving k∈{1, 2, …, K} users. In OFDMA-based LTE networks, the available bandwidth is divided into the number of resource blocks (RBs) based on the time and frequency domains^[23]. An RB is the smallest schedulable unit that can be modulated independently in an LTE network. An RB consists of 12 continuous subcarriers in the frequency domain, and has a 0.5 ms duration in the time domain. The allocation period is called one TTI, and it lasts for 1 ms duration. Therefore, the femtocell has the advantage of scheduling resources orthogonally to the UEs, thereby minimizing performance degradation due to interference.

In the system considered with a hybrid-access mode, if the UE has a high signal-to-interference-plus-noise ratio(SINR)from the nearest femtocell than from the originally associated macrocell, then the macrocell eNodeB can offload the UE to the femtocell, thereby increasing the overall cell capacity. However, the subscribers of femtocells are given priority for resource allocation when they are present in the coverage area of their registered femtocells. Nonsubscribers will be in connection with the macrocell eNodeB. When they are near the vicinity of the femtocell network, the connectivity of nonsubscribers should be assessed to determine whether they need to switch over to the femtocell network or if they remain connected with the macrocell eNodeB. This decision can be made by analyzing the link conditions between them and the base stations. However, they can be connected to the femtocell based on their service requirement and femtocell resource availability.

Nowadays, multimedia services such as voice over Internet protocol(VoIP)and video are becoming very important applications in telecommunication technology. These high-priority traffic technologies have strict delay and packet-loss requirements. When a user with these types of services is moving away from an eNodeB, the quality of experience of that particular user decreases. Femtocells are introduced mainly to improve the signal quality and to increase the overall cell capacity. Closed-access femtocells provide service only to registered subscribers, even though they have resources at a particular instant in time. These types of femtocells are generally preferred in home environments. However, open-access femtocells can serve all users based on their resource availability, but they are not preferred in home environments. Hybridaccess femtocells also share some limited resources with nonsubscribers. If femtocells serve users based on the traffic type, then the overall throughput of high-priority traffic can be improved. In the proposed method, the user who needs a handover from a macrocell to a femtocell is also included in the high-priority traffic. Because the proposed method uses hybrid access femtocells, the serving-station selection needs to be performed before resource allocation. The choice of serving station is made by the local entity, which performs functions such as link management as well as handover mechanisms between eNodeBs and HeNodeBs.

The serving station selection is performed by investigating the SINR of a particular user with respect to the macrocell eNodeB and femtocell HeNodeB. The SINR of a user at the nth RB with respect to the macrocell eNodeB is calculated as

where P_{m, n} is the transmission power from the macrocell eNodeB, G_{m, n} is the channel gain from the macrocell to the user, P_{f, n} is the transmission power of interfering femtocells, G _{f, n} is the channel gain from the femtocell to the user, N₀ is the noise spectral density, and w is the bandwidth of the RB. The SINR of a user at the nth RB with respect to the femtocell HeNodeB is calculated as

where P_{k, n} is the transmission power from the femtocell HeNodeB with k∈F, G_{k, n} is the channel gain from the femtocell to the user, P_{i, n} is the transmission power of interfering femtocells, and G_{i, n} is the channel gain from the interfering femtocells to the user. Initially, the user will calculate the SINR value for all the serving stations, including macrocell and femtocells, based on the channel quality it experiences. The serving station will adopt the corresponding modulation and coding scheme(MCS)for the user, as shown in Tab.1^[24], for its calculated SINR value.

After calculating the SINR values, the user will check if

where γ_M is the SINR from the macrocell eNodeB, γ_F is the SINR from femtocell HeNodeB, γ_th is the SINR threshold. This threshold value can be taken based on the operator’s requirement. For a simulation, 7 dB is taken as the threshold value. This is because with a threshold value above 7 dB, a user can switch from one modulation order to another that reduces the resource requirements for that user. If the user has an SINR difference that exceeds the threshold value, then it may request access to the femtocell. Eq. (3)is calculated only for nonsubscribers. The femtocell can grant access to the user based on the traffic type and resource availability. The steps involved at the local entity for connecting the UE with a femtocell or to make a handoff from a macrocell to a femtocell are shown in Algorithm 1.

Algorithm 1 At local entity

Calculate SINR of UE with respect to macrocell using Eq. (1)

Calculate SINR of UE with respect to femtocell using Eq. (2)

For every TTI do

if γ_M−γ_F> γ_th then

check the resource availability at femtocell

connect UE with femtocell or

else connect UE with macrocell

end if

end for

In hybrid-access mode femtocells, separate scheduling methods can be used for subscribers and nonsubscribers. Mostly, the conventional method of scheduling, which is called first-in first-out(FIFO), is adopted for nonsubscribers because of its simplicity. However, by giving preferen-tial access to the high-priority traffic of nonsubscribers, the overall throughput of the high priority traffic in the net-work is increased. High-priority traffic includes real-time services and handoff traffic from the macrocell, and low-priority traffic includes non-real-time services. To sched-ule nonsubscribers in femtocells, the service-based scheduler (SBS)method is adopted^[25]. The flowchart of the proposed resource-scheduling method in a hybrid-access femtocell is shown in Fig.3.

Initially, the resources are fixed for subscribers and nonsubscribers in the femtocell. If more resources are needed by the subscribers, then some resources are released from nonsubscribers’resource pool to guarantee the minimum rate requested by the subscribers in the femtocell. While performing resource allocation for nonsubscribers, preference is given to the high-priority services. For resource scheduling of the nonsubscribers, the SBS algorithm is used. The SBS algorithm estimates two values, namely T_HP and T_LP, for nonsubscribers based on their type of service, and they put nonsubscribers in a separate list. These lists are arranged in ascending order using any sorting algorithm, and the allocation of nonsubscribers is done initially from the T_HP list and T_LP then from the list. Based on the reported channel quality indicator(CQI)value obtained from the user, RBs are assigned. The values of T_HP and T_LP for a particular packet is calculated as follows:

where T_d is the theoretical departure time, which is the difference between the packet delay budget value and the head of line(HoL)delay value, $L_i^k$ is the packet size, and r_i(min)is the minimum guaranteed bit rate.

where r_i(max) is the maximum bit rate of that service. Using this method, high-priority traffic obtains preferential access, and the overall throughput of the high-priority traffic is improved. The steps involved in the femtocell for providing resources to nonsubscribers are shown in Algorithm 2.

Algorithm 2 At femtocell

While the nonsubscribers join the network do

Calculate T_HP for high priority UEs using Eq. (4)

Calculate T_LP for low priority UE using Eq. (5)

Sort T_HP and T_LP list separately

For every TTI do

if current TTI has resources

after allocating subscribers then

select UEs from T_HP list

allocate resources

if current TTI has resources

after allocating UEs from T_HP list then

select UEs from T_LP list

allocate resources

end if

end for

A single-cell DL scenario is considered with femtocells (the number ranges from 1 to 30) distributed randomly inside the cell without overlapping each other. It is assumed that users move with speeds of 3 km/h and 30 km/h. For outdoor users, a 6-ray typical urban channel model is adopted, and for indoor users, the WINNER Ⅱ channel model is used. To minimize the complexity of the simulation, the reception of a perfect CQI report from the user is assumed. It is also assumed that at any instance of time, 60% of users will be in high-priority traffic and 40% of users will be in low-priority traffic. High-priority traffic includes VoIP, video, and handoff users, while low-priority traffic includes BE service users. Video flows with a source rate of 128 kbit/s are produced by a H. 264 trace-based coder, whereas VoIP flows with a source rate of 8 kbit/s are generated by a G. 729 audio coder. Finally, BE traffic is generated by an ideal source that always has packets to send. The main parameters used in the simulation are shown in Tab.2.

The performance of the proposed resource-allocation method is evaluated by increasing the number of femtocells in the given macrocell region using an open-source simulator, LTESim. The modules present in LTESim are explained in Ref. [26]. The average throughput of high-priority users at 3 km/h and 30 km/h is shown in Fig.4, from which it is observed that the average throughput of high-priority users increases with an increasing number of femtocells owing to the effective utilization of system resources. Because users are chosen from the T_HP list for resource allocation, high-priority users are allocated first, thereby increasing their overall average throughput. However, as the user’s speed increases, there is a high probability of packet loss, which leads to a decrease in the average throughput that can be seen in all the figures with two different user speeds.

From Fig.4, it is also observed that by using the pro posed resource-allocation method for nonsubscribers, there is considerable improvement in the average throughput of highpriority users when compared with the conventional method for nonsubscribers. When the femtocells use the conventional resource-scheduling method for nonsubscribers, users are allocated based on their entry time. Based on the proposed priority-based resource-allocation method, preferential access is given to high-priority users. Thereby, the overall average throughput of high-priority users is increased. Using the proposed resource-scheduling method, there is a 6. 8% improvement in the average throughput of nonsubscribers when compared to the conventional scheduling method with the parameters assumed for simulation at 3 km/h. At 30 km/h, there is a 14% improvement in the average throughput with the parameters taken for simulation. At the same time, the delay of the high-priority users is decreased, as shown in Fig.5, because of the preferential access to those users. There is a 5% and 4. 5% decrease in the average delay at 3 km/h and 30 km/h, respectively with the parameters assumed for simulation.

At the same time, a compromise should be made in the average throughput and delay of low-priority users, as shown in Fig.6 and Fig.7. There is a 7. 3% decrease in the average throughput and a 7% increase in the average delay at 3 km/h, and a 12% decrease in the average throughput and a 6. 1% increase in the average delay at 30 km/h for low-priority users with the parameters assumed for simulation. The increase in the delay with an increase in the number of femtocells shows the impact of interference in the network.

The average throughput of nonsubscribers when compared with other access modes is shown in Figs. 8 and 9. In general, the performance of nonsubscribers is higher in the hybridaccess mode, which is confirmed from the figures. The reason is that primarily, the cell-edge users who suffer from poor signal quality will try to access the femtocell network. With the hybrid-access method, these users are effectively served, increasing the average throughput of nonsubscribers. However, open-access femtocells allow all the users that are within its coverage area, increasing the number of handovers. Therefore, the average throughput of the open-access method is less than that of the hybrid-access method. For the same environment, the closed-access method gives access only to subscribers, and nonsubscribers therefore need to be connected only to the macrocell even though they have to communicate with poor signal quality. Thus, the average throughput of nonsubscribers in the closed-access mode is significantly less when compared with other access methods.

The performance of the average throughput of all users with different access methods is shown in Figs. 10 and 11. In general, the performance of the closed-access method is superior in this case when compared to other access methods, which is confirmed by figures. This is because with the closedaccess method, the subscriber’s individual average throughput is maximized such that it improves the overall average throughput of all users in the network. However, in the hybridaccess mode, the subscribers’minimum throughput is guaranteed to serve a limited number of nonsubscribers, which improves the overall average throughput of all users. But it is less compared to the closed-access method. The open-access method shows poor throughput performance owing to an increase in signaling and the number of handovers.

When compared with the conventional scheduling method for nonsubscribers, the proposed method shows better performance in terms of the average throughput of all users. The conventional method treats all the nonsubscribers equally, such that it increases the delay experienced by high-priority services. When the number of femtocells increases, the impact of delays in the network also increases, reducing the overall throughput of users in the network, as shown in Fig.12. Using the proposed priority-based resourcescheduling method for nonsubscribers, there is an 8. 2% improvement in the average throughput of total users when compared to the conventional scheduling method at 3 km/h, as obtained with the parameters assumed for the simulation. At the same time, there is a 15% improvement in the average throughput at 30 km/h, as obtained with the parameters assumed for the simulation.

This study investigates resource scheduling for femtocell networks in the hybrid-access mode, which gives preferential access to high-priority users in a network. Because the proposed method uses hybrid-access femtocells, the effective serving-station selection is performed before resource allocation. This serving-station selection is done by the local entity of the network based on the individual users’SINR feedback. After the serving-station selection, resource schedul ing is done by giving preferential access to the nonsubscribers that have high-priority traffic. It is performed by adopting the already-existing SBS algorithm. By estimating T_HP and T_LP values and allocating the users, initially from the T_HP list and then from the T_LP list, the overall throughput of the highpriority users is increased. Using the priority-based resourcescheduling method, there is a 6. 8% and 14% improvement in the average throughput of high-priority users at 3 km/h and 30 km/h, respectively, with the parameters assumed for the simulation. At the same time, the delay is reduced by 5% and 4. 5% at 3 km/h and 30 km/h, respectively.