A fog-driven dynamic resource allocation technique in ultra dense femtocell networks

https://doi.org/10.1016/j.jnca.2019.102407Get rights and content

Abstract

The highly dense small cell structure filled with large number of Femtocell Base Stations (FBSs) is expected to address the increasing data demand of end users in current and upcoming generation of wireless networks. However, the large and random deployment of such devices incur severe interference which leads to significant performance degradation. To overcome this issue in the Orthogonal Frequency Division Multiple Access (OFDMA)-based femtocell networks, we propose a hierarchical technique consisting of a dynamic distributed clustering and a fog-driven resource allocation to optimize the total throughput of the network while mitigating the interference. Our fully distributed clustering method is designed so that FBSs adaptively form clusters with dynamic size based on the current status of the network and end users. Moreover, we put forward a policy-aware resource allocation method to address the intra and inter-cluster interference, which are two potential types of interference in clustering-based resource allocation techniques. Since our technique carefully considers users' demands in cluster formation, there is always sufficient resources for end users in each cluster, so that each cluster head can find a resource allocation solution, by which no intra-cluster interference occurs. Besides, we employ local fog servers situated in the proximity of clusters for monitoring and assigning a set of policies to CHs for resource allocation, by which the number of inter-cluster interference can be significantly reduced. The extensive simulation results demonstrate that our proposed hierarchical technique significantly improves total throughout, interference, user satisfaction, and fairness compared to other proposed techniques in dense and ultra-dense femtocell networks.

Introduction

A rapid growth in deployment and the use of mobile devices such as smartphones, tablets, and sensors has resulted in rapid increase of data-streaming applications such as video streaming, online games, health-care, and Voice over Internet Protocol (VoIP). This leads to a significant amount of data to be transferred over cellular networks (Lee et al., 2014; Goudarzi etal., 2016, Goudarzi etal., 2017). Considering the fact that the number of cellular network resources is restricted, the requested quality of service can be satisfied for only a limited number of users. Besides, recent studies have revealed that approximately 70 percent of the data is originated from indoor places where severe wall penetration loss and longer transmission distance incur poor received signal quality (Garcia-Morales et al., 2015). To address these issues, Femtocell Base Stations (FBSs), which are low-power, short-range, and low-cost edge devices are deployed over macrocell network to effectively improve the indoor received signal quality and overall network throughput. This latter is obtained by reusing same frequency by several FBSs while the former one is satisfied by decreasing the distance between transmitter and end users (Fu et al., 2017).

However, in densely deployed femtocell networks, neighboring FBSs experience severe co-tier interference (i.e., interference between adjacent femtocells (Mhiri et al., 2013)) due to finite domain of shared spectrum unless an efficient interference management technique is used. The co-tier interference can be significantly reduced in downlink Orthogonal Frequency Division Multiple Access (OFDMA)-based femtocell networks by means of an efficient allocation of Resource Blocks (RBs) between interfering FBSs (Bu et al., 2015). To achieve this, researchers have proposed several Resource Allocation (RA) techniques including centralized and clustering. However, due to non-convex non-deterministic polynomial time (NP-hard) nature of this problem, centralized techniques are not efficiently practical and result in high complexity, signaling overhead, and single point of failure, specifically in dense and ultra-dense networks (Fu et al., 2017; Rohoden et al., 2019). To overcome this problem, the clustering-based RA techniques, which are partially decentralized, are introduced by which the complexity of RA problem is significantly reduced. In majority of these techniques, each cluster has access to entire set of RBs, while FBSs in one cluster cannot use the same RBs simultaneously. This latter enables RA technique to be performed in each cluster independently of other clusters (Lee et al., 2014).

In order to effectively utilize the benefits of clustering in RA, several issues should be carefully addressed. Clusters can be formed either by gateway (GW) centrally or by FBSs in a distributed manner (Qiu et al., 2016). Moreover, the maximum size and number of clusters can be statically determined or can be obtained dynamically by the GW or cluster heads (CHs) at the runtime. In addition, the RA in each cluster can be performed by a CH individually or all FBSs collaboratively. Besides, in the dense and ultra-dense femtocell networks, interference between clusters should be mitigated so that FBSs located at the edge of clusters (edge FBSs) do not suffer from decreased throughput, which apparently reduces total throughput and end users' quality of experience. Last but not least, it is worth mentioning that centralized and clustering-based RA techniques, in which GWs and CHs respectively perform the majority of responsibilities, suffer from the scalability issues in dense and ultra-dense femtocell networks, because the above-mentioned burdens are not proportionally distributed.

Considering the aforementioned issues, we propose a Distributed Dynamic Clustering (D2C)-FOg-driven Resource Allocation Technique (D2C-FORAT) to optimize the total throughput of the downlink OFDMA femtocell networks. The proposed solution is divided into two methods including distributed dynamic clustering and RA, so that we proportionally distribute responsibilities over the network entities including FBSs, CHs, GW, and local fog servers. The fog servers are local entities located in the proximity of end users, which have computing capabilities, and can be accessed by low latency (Zhou et al., 2016; Hu et al., 2017; Chang et al., 2019). In the D2C-FORAT, FBSs make clusters in a distributed dynamic manner so that FBSs which have the highest co-tier interference on each other join to the same cluster, and select a CH. Afterward, the CH monitors the available resources and users' demands in its cluster, and dynamically control the size of its cluster in the runtime. In addition, the fog servers collect the edge FBSs' information of each cluster which is then used to form the edge FBSs' interference graph. Besides, the fog servers employ a graph-coloring-based technique to assign a set of policies for edge FBSs in each cluster to reduce the inter-cluster interference. These policies are then forwarded to respective CHs, by which the RA can be performed more efficiently, resulting in increased throughput and user satisfaction.

We summarize the main contributions of this paper as follows.

  • 1)

    We propose a hierarchical RA technique, aiming at maximizing the total throughput while mitigating the interference, to satisfy the ever-increasing users' demands in dense and ultra-dense femtocell networks

  • 2)

    We put forward a distributed dynamic clustering algorithm by which CHs adaptively control their cluster size based on requested demands of their end users. This results in better scalability so that our technique can be effectively adapted to dense and ultra-dense femtocell networks.

  • 3)

    Considering the fact that sufficient resources are available in each cluster due to our clustering method, no intra-cluster interference occurs. To address the inter-cluster interference problem, we develop a fog-driven RA method by which the fog servers assign a set of policies to CHs to be considered in their RA. This latter leads to decreasing the inter-cluster interference which significantly improve the total throughput and user satisfaction.

  • 4)

    We study current clustering-based RA techniques in femtocell networks to identify their key elements, and provide a comprehensive qualitative comparison.

  • 5)

    The performance of our technique, D2C-FORAT, is comprehensively evaluated in dense and ultra-dense femtocell networks, and we compared it by the state-of-the-art current techniques in terms of system throughput, interference, satisfaction rate, and fairness to precisely analyze its efficiency.

The rest of the paper is organized as follows. Section 2 reviews the current literature in clustering-based RA techniques in femtocell networks. The system model and problem formulations are presented in section 3. Our distributed clustering and RA methods are presented in section 4 and section 5, respectively. In section 6, we evaluate the system performance under our proposed solution, and compare it by the state-of-the-art RA techniques of the literature. Finally, section 7 concludes the paper and draws future works.

Section snippets

Related work

A significant number of studies has been focused on RA techniques in OFDMA-based femtocell networks to address the co-tier interference, among which we study the current literature in clustering-based RA. The proposed techniques are categorized into two groups of centralized and distributed based on their clustering approach. Besides, main elements of each technique are identified, by which we can qualitatively compare these techniques.

In the centralized clustering techniques, clustering is

System model and formulation

In this section, we describe the system model and formulate the RA as an optimization problem to maximize the network throughput.

Distributed dynamic clustering method

In this section, we propose a Distributed Dynamic Clustering (D2C) method, in which FBSs with highest relative interference form different clusters. Moreover, FBSs in each cluster select one FBS as their CH. The CH dynamically controls the cluster size based on requested demands of its end users and makes the decision whether a new FBS can join the cluster or not, accordingly. The D2C has three principal functions including new FBS arrival (NFA), update clustering parameters (UCP), and cluster

A new resource allocation method

In this section, we propose a RA method in which fog servers and CHs collaborate to mitigate the interference and improve the network throughput.

Each CH is responsible to allocate the RBs so that no intra-cluster interference occurs in its cluster. Since each CH is unaware of adjacent clusters' RAs, there is high probability of inter-cluster interference on edge FBSs. This problem is aggravated in dense and ultra-dense networks to the point that the network throughput is severely dropped (Qiu

Performance evaluation

In this section, we evaluate the performance of our proposed solution through extensive simulations under different scenarios and compare it by the state-of-the-art RA techniques in femtocell networks to understand its efficiency. We discuss the employed system parameters and study the obtained results in the performance study subsection.

Conclusions and future work

In this work, we proposed a distributed dynamic clustering-fog driven RA technique, called D2C-FORAT, to address the interference problem of femtocell networks, and to increase the total network throughput. Moreover, we used a hierarchical architecture, including the GW, fog servers, CHs, and CMs, among which the clustering and RA responsibilities are distributed. This latter results in better scalability, helping our technique to be efficiently run in sparse, dense, and ultra-dense networks.

Conflict of interest

No conflict of interest.

Acknowledgments

We thank Sara Kardani Moghaddam for her comments on improving this paper. We also thank anonymous reviewers and editors for their valuable feedback in enhancing the quality of our paper.

Mohammad Goudarzi graduated from Iran University of Science and Technology (IUST), Tehran, Iran, with First-Class Honors degree in M.Sc. in Information Technology (Computer Networking), where he was selected as the exceptional talented student as well. Due to his academic achievements, he was awarded to become a member of Iranian National Elites Foundation, a prestigious organization for recognition and support of Iranian national elites, from which he received a prestigious research Grant. He

References (30)

  • J. Garcia-Morales et al.

    Analytical performance evaluation of ofdma-based heterogeneous cellular networks using ffr

  • M. Goudarzi et al.

    A genetic based decision algorithm for multisite computation offloading in mobile cloud computing

    Int. J. Commun. Syst.

    (2016)
  • V.N. Ha et al.

    Fair resource allocation for ofdma femtocell networks with macrocell protection

    IEEE Trans. Veh. Technol.

    (2014)
  • A. Hatoum et al.

    Cluster-based resource management in ofdma femtocell networks with qos guarantees

    IEEE Trans. Veh. Technol.

    (2014)
  • R.K. Jain et al.

    A Quantitative Measure of Fairness and Discrimination

    (1984)
  • Cited by (0)

    Mohammad Goudarzi graduated from Iran University of Science and Technology (IUST), Tehran, Iran, with First-Class Honors degree in M.Sc. in Information Technology (Computer Networking), where he was selected as the exceptional talented student as well. Due to his academic achievements, he was awarded to become a member of Iranian National Elites Foundation, a prestigious organization for recognition and support of Iranian national elites, from which he received a prestigious research Grant. He is working towards the Ph.D. degree at the Cloud Computing and Distributed Systems (CLOUDS) Laboratory, Department of Computing and Information Systems, the University of Melbourne, Australia. He was awarded the Melbourne International Research Scholarship (MIRS) supporting his studies. Besides, he was awarded the Rowden White Scholarship, a prestigious scholarship provided by the University of Melbourne to talented, high quality PhD students. His research interests include Internet of Things (IoT), Fog Computing, Wireless Networks, and Optimization.

    Marimuthu Palaniswami (F'12) received the M.E. degree in electrical, electronic, and control engineering from the Indian Institute of Science, Bengaluru, India, in 1979, the M.Eng.Sc. degree in electrical, electronic, and control engineering from The University of Melbourne, Melbourne, VIC, Australia, in 1983, and the Ph.D. degree from The University of Newcastle, Callaghan, NSW, Australia, in 1987. He is currently a Professor in electrical engineering and the Director/Convener of a large ARC Research Network on Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP) with about 100 researchers on various interdisciplinary projects. He is representing Australia as a core partner in EU FP7 projects, such as SENSEI, SmartSantander, Internet of Things Initiative, and SocIoTal. He is the author or coauthor of more than 480 refereed research papers and leads one of the largest funded ARC Research Network on ISSNIP. His research interests include smart sensors and sensor networks, machine learning, IoT, and biomedical engineering and control.

    Rajkumar Buyya is a Redmond Barry Distinguished Professor and Director of the Cloud Computing and Distributed Systems (CLOUDS) Laboratory at the University of Melbourne, Australia. He is also serving as the founding CEO of Manjrasoft, a spin-off company of the University, commercializing its innovations in Cloud Computing. He has authored over 650 publications and seven text books including “Mastering Cloud Computing” published by McGraw Hill, China Machine Press, and Morgan Kaufmann for Indian, Chinese and international markets respectively. Dr. Buyya is one of the highly cited authors in computer science and software engineering worldwide (h-index = 124, g-index = 271, 79,100 + citations). He is named in the recent Clarivate Analytics' (formerly Thomson Reuters) Highly Cited Researchers and “World's Most Influential Scientific Minds” for three consecutive years since 2016. Dr. Buyya is recognized as Scopus Researcher of the Year 2017 with Excellence in Innovative Research Award by Elsevier for his outstanding contributions to Cloud computing. Software technologies for Grid, Cloud, and Fog computing developed under Dr.Buyya's leadership have gained rapid acceptance and are in use at several academic institutions and commercial enterprises in 40 countries around the world. Manjrasoft's Aneka Cloud technology developed under his leadership has received “2010 Frost & Sullivan New Product Innovation Award”. He served as founding Editor-in-Chief of the IEEE Transactions on Cloud Computing. He is currently serving as Editor-in-Chief of Software: Practice and Experience, a long standing journal in the field established ∼50 years ago. For further information on Dr.Buyya, please visit his cyberhome: www.buyya.com.

    View full text