Performance Evaluation of IPTV Zapping Time Reduction Using Edge Processing of Fog RAN

The shifting towards Internet-based services is one of the key trends in the present and future technologies [1]. This is coming side by side with the noticeable convergence between IP communication and broadcasting. Internet Protocol Television (IPTV) system has gained a particular interest recently to provide ubiquitous TV service delivery, via the Internet as a transmission medium. This means that a full package of services can be provided by the Internet including surfing the web, free internet-based mobile phone calls, and the provision of TV channels are also included altogether to the end-users [2]. One of the open research problems in the IPTV system is the challenge of zapping time when the user tries to switch from one channel to another. In IPTV system, the delay may extend to several seconds, which requires lots of effort from the service providers to meet the user's satisfaction with the provided service [3], and this is commonly known as the Quality of Experience (QoE). Hence, QoE is a key metric that measures the success of the IPTV system [4]. Unlike the Quality of Service (QoS), which is exclusively related to measurements of networking parameters (e.g., packet loss, jitter, and delays), the QoE is a measure of the level of customer satisfaction with the service offered by vendors. Hence, there is a positive impact when customer's experience with high quality TV channels that display “immediately” on screen using the IPTV service. Prompt response in IPTV system can be considered as a competitive factor between operators. However, the zapping delay between two successive channels may lead to inconvenience by the end users.

Zapping time may be an obstacle to the expansion of IPTV system service. Unlike the traditional broadcasting TV system which can display the channel content in a small time, the receiver device of the IPTV technology, cannot receive all channels simultaneously over the IP-based network due to bandwidth limitation [5, 6]. On the other hand, the IPTV system presents a promising technology to provide smart services such as detecting the essential channels based on the preference and the behavior of the users [7].

Fog Radio Access Networks (F-RAN) has been developed as a promising cellular architecture that can meet the requirement of future low communication latency applications [8]. Hence, F-RAN architecture is proposed in this work to examine the possibility of reducing the Zapping Time (ZT) via utilizing the property of edge computing in F-RAN architecture. The remainder of the paper is organized as follows: in section 2, a concise review of the prior work is demonstrated. In Section 3 a theoretical explanation for the IPTV system, zapping time, along the F-RAN architecture are presented. In Section 4, the proposed structure of the IPTV/F-RAN is demonstrated. Section 5 illustrates the evaluation and the discussion. Finally, Section 7 contains the conclusion of the paper.

Throughout this section, the main components of this research are explained and clarified including the IPTV system, zapping time problem, and the evaluation of RAN in cellular networks.

3.1 Mobile internet protocol television wireless networks

In the earlier time of the IPTV system, the service has been provided without mobility using a wired access network with a special device known as Set-Top Box (STB) which is co-located with TV sets. Later, the continued development in wireless technology has paved the way to access the Internet from anywhere at any time.

Table 1. A comparison between the legacy wired-based and the wireless access network of IPTV system [20]

Table 1 presents a brief comparison between the traditional and wireless access IPTV system. IPTV represents an integration of television services to high-speed networks. This enables IPTV to support a seamless wireless communication environment to provide TV channels in real-time over IP networks. Wi-Fi is the most common technology that has been used with IPTV systems over home networks. However, employing Wi-Fi as a shared wireless home network can form a bottleneck point particularly with the multimedia applications that require high bandwidth demands. Hence, the cellular network is a promising technology that can be utilized to provide ubiquitous live broadcast services (TV channels) at the high Quality of Service (QoS), quality of experience, security, and reliability [21]. The main challenges of the IPTV system are the rate of packet loss; packet jitter; the end-to-end delay and the channel zapping time [22].

3.2 Zapping delay time problem

Channel zapping time occurs when the user switches from one TV channel to another as shown in Figure 1. It is worth stating that in the traditional cable-based TV, channels are received simultaneously with a very short zapping time without a noticeable delay. However, in the Internet-based platform, several factors or limitations may lead to generating zapping time around (0.9-70) seconds throughout switching of channels [23]. Consequently, high delays lead to users unsatisfactory with the IPTV system. Hence, zapping time is one of the key factors that affect QoE regarding the service of the IPTV system perceived by subscribers.

1.jpg

Figure 1. Illustration of channel zapping time and viewing time

There are several sources of zapping time delay. The key components of this delay as illustrated in [14, 24-26] including 1) the Internet Group Management Protocol (IGMP) that is responsible for sending messages through the network to join the new channel and leave the previous one; 2) the synchronization delay which is described as the time between the reception of reference I-frame at the STB and the beginning of the decoding process; 3) buffering delay, which is the time of buffering video frames. It is worth stating that the operation of IGMP communication protocol is presented by establishing a multicast group membership for IP based networks of online streaming videos. In other words, IGMP is used to direct the multicast transmission to the hosts based on their requests. Hence, the delay that is induced in the network during the IGMP transmissions can be referred theoretically as a network delay. Regarding the theoretical discerption of zapping delay time in the IPTV system, the authors in [27] have describing mathematically the zapping time via dividing the delay into three main parts as shown in Eq. (1) as follows:

$Z_T=D T_s+D T_b+D T_N$    (1)

where, Z_T is channel zapping delay time, DT_s is the average synchronization delay, DT_b is the buffering delay, DT_N is the network delay, which includes IGMP delay.

$D T_s=\frac{\delta}{2 \theta}$   (2)

where, θ is no. of frames per second for the broadcasted videos, δ is the no. of frames within the group of pictures.

$D T_b=\frac{\mu \delta}{\theta}$   (3)

where, μ is the essential group number of pictures in the receiver to begin playing.

$D T_N=\sum_{k=1}^m p_i \tau_i$  (4)

where, p_i is the probability to join the group of multicasts at node k, τ_i is the time duration from the transmission of the request of video chunks until receiving the video chunk, m is an index for the sender.

3.3 Evolution of radio access network from dedicated hardware to cloud and fog cloud

In a mobile network, the Radio Access Network (RAN) has been evolved chronologically definitely to meet the new demands of cellular networks such as the massive number of subscribers, the QoS provision and at the same time to increase the revenue of mobile network operators. Starting from the 4G LTE Distributed RAN (D-RAN) network architecture which represents the point of change towards the virtualization. In LTE/LTE-A, the RAN includes a physical base station (eNodeB) connected via the fronthaul communication link to the access points or the Remote Radio Heads (RRHs). However, this paradigm has several challenges such as high deployment costs, cell interference management, high energy consumption, and traffic load management, for more details, the interested reader may refer to [28, 29].

In the C-RAN architecture [30], a new concept has been introduced to cellular networks, which is the virtualized RAN that is moved into cloud computing to utilize the features of the cloud such as processing and storage capabilities. Interestingly, C-RAN can solve most of the challenges of the traditional hardware-based or known as distributed RAN network. C-RAN, however, also has some challenges in the deployment including the high burden on the fronthaul communication link in terms of capacity and latency constraints which leads finally to decrease the spectral efficiency. Hence, to overcome of the challenges of C-RAN, the Heterogeneous C-RAN (H-CRAN) architecture has been developed [31] to decouple the control and user planes.

H-CRAN unfortunately still has disadvantages [7] represented by first the operators have to use a high number of RRHs to meet high-capacity requirements. In addition to a low gain is achieved from the storage and processing of the edge devices RRHs and smart User Equipments (UEs) to lighten the load between the fronthaul and the virtual BBU. Besides, the traffic data between RRHs the pool of BBUs can include redundant information that affects the constraints of the fronthaul. To solve the aforementioned challenges, the concepts of Fog-RAN have been proposed as revolutionary architecture [7, 32]. F-RAN is an architecture that brings some of cloud computing capabilities from the center to the edge of the network including communication, storage, control configuration, and management. This evaluation brings several new features and services, such as support of low latency applications, mobility, and a huge figure of nodes with position awareness. F-RAN is alternative solution for signal processing when the cloud providers have no data center at certain locations [33]. F-RAN supports wireless services and it can integrate with the present and future cellular network. Figure 2 illustrates the three key architectures with the manner of signal processing.

2.jpg

Figure 2. Distributed versus centralized signal processing of common cellular architectures

This section presents a concise analysis of the key aspects that can improve the network performance under the deployment of F-RAN, which can lead to minimizing the ZT of the IPTV system. For the network model, it is assumed that the F-RAN network is deployed with, N_i number of UEs (i=1, 2, …, N_i), which are served by M_j number of F-APs (j=1, 2, …, M_j) all have similar storage capacity in terms of local caches. The evaluation scenario is conducted between the C-RAN architecture with the F-RAN as illustrated in Figure 7.

The enhancement in the deployment of F-RAN is analyzed based on two points of views, including the communication concept along with the processing capabilities.

7.jpg

Figure 7. Flowchart of the proposed evaluation procedure

In the part of communication analysis, the Signal to Noise Ratio (SNR), the path loss, the received power, and the average link throughput are investigated.

For a single-hop wireless transmission between the UE and the F-AP (which represents an access point with light storage and processing capabilities), the SNR formula can be calculated as follows [36]:

$S N R_i=\frac{P_S K 10^{\frac{\delta_s}{10} \varphi}}{d^{\propto}\left(\sum_1^J P_j+T h_{U E}\right)}$       (5)

where, P_s is the transmission power, P_j is the interference power, d represents the distance between the UE and the F-AP, K and $\alpha$ are constant parameters inferred from propagation loss, K is Boltzman constant, L_tr is the penetration loss, ∝ equals 3.5, Th_UE is the thermal noise of the UE, $\varphi$ is a standard random variable for modeling the fading, δ_s is the standard deviation of the shadow fading.

According to Eq. (5), the SNR is inversely proportional to the value of the distance d. Hence, theoretically with the deployment of F-RAN, the SNR will be improved and this leads to improvement in the channel capacity as shown in the following expression [37].

$C=\sum_{H N} \sum_{C H} B W_i \log _2\left(1+S N R_i\right)$    (6)

where, BW_i is the bandwidth assigned to channel (i). Equation (6) shows that the overall capacity C corresponds to the summation of all Heterogeneous Networks (HN) cells and all their corresponding Channels (CH). It is worth stating that the most common routes that can be followed to increase the network capacity are (i) enlarging the number of access points of different types (heterogeneous), which is the core idea of using Fog-RAN with the dense deployment of Fog nodes; (ii) increasing the total number of channels; (iii) increasing the system bandwidth such as using higher frequency band in centimeter and millimeter waves [38], and finally (iv) increasing the SNR such as bringing the F-AP closer to the users. The next term considered is the wireless channel path loss. From a theoretical point of view, shorter distance means lower path loss and vice versa as illustrated in Eq. (7) [39]. Figure 8 describes the correlation between the path loss and the corresponding distance. It can be noticed that as the distance increases in the communication link, the attenuation in the radio frequency wave is increased as well. Likewise, the received power based on the Friis formula in Eq. (8) will be higher as the distance decreases (Inverse relationship) as shown in Figure 9.

$P L=92.4+20 \log 10(d)+20 \log 10(f)+\alpha(d)+\rho(d)+\epsilon+\delta_s$      (7)

where, PL is the path loss, d distance between the user and the F-AP, f is the frequency, ρ is the rain attenuation, ϵ is the Foliage losses, and δ_s is the standard deviation of the shadow fading.

The received power can be expressed as follows.

$P_r=P_t G_t G_r\left(\frac{c}{4 \pi d f}\right)^2$        (8)

where, P_r is the received power, P_t is the transmit power, G_t & G_r represent the transmitter and the receiver gains respectively, and c is the speed of light.

In fact, using the F-RAN can significantly improve the network performance via utilizing the merits of this architecture. These merits are thoroughly explained in [30]. The latency minimization is at the forefront of the advantages of the F-RAN architecture. The applications of fog end nodes not only reduce the communication latency but also minimize the fronthaul load traffic effectively since there is no need to pass through a long transmission network to the remote core network for processing. According to [40-42], in F-RAN, with the dense deployment of edge fog nodes, the user can access the F-AP’s which are nearby. Unlike the C-RAN, in F-RAN, the latency that may generate from the propagation delay as well as from the centralized computational overhead will be significantly reduced due to the decentralized management [43]. Improving SNR in F-RAN architecture increases channel capacity as shown in Eq. (6) which consequently leads indirectly to minimizing the zapping delay. In fact, the low network capacity means that the incoming data may reach the network capacity and this increases the delay sharply by adding waiting time for the queues of all nodes in the path, which is known as network congestion. Hence, in the F-RAN scenario the overall network delay τ_i shown in Eq. (2) will be decreased significantly. The second part of analysis of ZT with F-RAN architecture is represented by the distributed signal processing over the edge nodes instead of full centralized processing as in C-RAN. The advantageous aspect of employing the F-RAN architecture in the IPTV system can be analyzed using the concept of divide and conquer algorithm as demonstrated in Algorithm 1. This is due to fact that the operation nature of the F-RAN architecture is relied on the concept of dividing the processing of large tasks that had been resulted previously from huge number of UEs as in the centralized processing paradigm of C-RAN to be in form of distributed chunks via allocating a specific number of UEs to each server (edge node). Hence, the edge node can deal with the its edge UEs only.

Regarding the time complexity of a divide-and-conquer algorithm can be expressed as follows:

$T(n)=\left\{\begin{array}{cc}T\left(\frac{n}{k}\right)+f(n)+c, & n>\text { Min_}_{-} \text {Size_Threshould } \\ 1 & , n \leq \text { Min_Size_Threshould }\end{array}\right.$      (9)

where, n: size of input, c: is a real constant. $\left(\frac{n}{k}\right)$: size of each subproblem. All subproblems are assumed to have the same size each requires $T\left(\frac{n}{k}\right)$ time. f(n): cost of the work done outside the recursive call, which includes the cost of dividing the problem and cost of merging the solutions.

For parallel Divide-and-Conquer Algorithms, the function of complexity of the divide-and-conquer algorithm can be determined in O(logn) parallel time. The operation of F-RAN architecture is compatible with the concept of Divide-and-Conquer Algorithms since it enables the network to divide the signal processing tasks into several servers (fog nodes). This can be achieved via moving the computation from the cloud into the fog nodes to manage the processing of a delay sensitive applications. This can be expressed mathematically using M/M/C queueing system as the $\rho=\frac{\lambda}{N \mu}$ instead of $\rho=\frac{\lambda}{\mu}$, where, ρ is traffic intensity (utilization factor), λ is the arrival rate, μ is the service rate and N is number of parallel servers.

8.png

Figure 8. Path loss versus distance

9.png

Figure 9. The trade-off between distance and power