Implementing caching to ultra-densely deployed small cells provides a promising solution for satisfying the stringent quality of service (QoS) requirements of delay-sensitive applications with limited backhaul capacity. With the rapidly increasing energy consumption, in this chapter, the authors investigate the NP-hard energy-efficient context-aware resource allocation problem and formulate it as a one-to-one matching problem. The preference lists in the matching are modeled based on the optimum energy efficiency (EE) under specified matching, which can be obtained by using an iterative power allocation algorithm based on nonlinear fractional programming and Lagrange dual decomposition. Next, on account of the Gale-Shapley algorithm, an energy-efficient matching algorithm is proposed. Some properties of the proposed algorithm are discussed and analyzed in detail. Moreover, the authors extend the algorithm to the matching with indifferent and incomplete preference lists. Finally, the significant performance gain of the proposed algorithm is demonstrated through simulation results.
TopI. Introduction
It is predicted that the connected devices will reach almost 50 billion by 2020, and the corresponding mobile data traffic is expected to grow more than 1000 times (Boccardi et al., 2014; Bello & Zeadally, 2014). Since acquiring new spectrum is becoming increasingly difficult due to the physical limits of the spectral efficiency (SE), it is extremely challenging for the current long term evolution (LTE) system to meet the demands of increasing capacity by another 1000 times (Li, Niu, Papathanassiou, & Wu, 2014).
One promising approach when solving the above challenge is to deploy a large number of small cells (SCs) within the legendary macro cell network for further SE enhancement by reducing the distance between users and contents (Shanmugam, Golrezaei, Dimakis, Molisch, & Caire, 2013). Compared with traditional large-scale, high-cost macro cell base stations (MBSs), small cell base stations (SCBSs) with the advantage of lower power consumption and operational cost enable high-density spectrum reusing and short-range communications (Shanmugam et al., 2013). However, several resource allocation challenges remain to be addressed for the ultra-dense deployment of SBSs along with the centralized MBS. For delay-sensitive applications, high-speed backhaul links are an essential requirement for reliable quality of service (QoS) provision. When the cell density and user density are of the same order of magnitude, the cost to equip the ultra-densely deployed SBSs with high-speed backhaul links will be enormously high (Shanmugam et al., 2013; Rusek et al., 2013). To address the above challenges, caching popular contents at the edge of the network provides an efficient solution for SBSs with limited capacity backhaul links to accommodate highly overloaded traffic. In this way, the lost-cost storage capacity can be considered as an effective substitute for the high-cost backhaul capacity.
In this chapter, we will firstly introduce the related works and the state-of-the-art progress in the direction of edge-caching based resource allocation. There has been extensive research on the resource allocation problems in edge-caching based SC networks. These works mainly aimed at maximizing the SE in the process of resource allocation regardless of the energy efficiency (EE). The solutions proposed in these works are not easy to be extended for optimizing EE since it is not always possible to realize the maximum EE and SE simultaneously (Chan, Zhang, Xu, & Li, 2011). In addition, EE has already become a crucial factor for the performance of cellular networks with rapidly increasing energy consumption (Chan et al., 2011; Bu, Yu, Cai, & Liu, 2012). The annual operational expenditure (OPEX) of mobile network in terms of electricity has already reached 10 billion US dollars, and the electricity expenditure of BSs accounts for almost 60%-80% (Oh, Krishnamachari, Liu, & Niu, 2011). The amount of CO2 generated by cellular networks is in the same order of magnitude of 8 million vehicles (Bu et al., 2012). In the future, we can predict that the energy consumption and CO2 emission will experience a significant increase with the ultra-dense deployment of a large number of SBSs and their relevant supporting infrastructures. Therefore, the focus of this chapter is placed on how to design an energy-efficient context-aware matching approach for resource allocation by exploiting properties of matching theory, nonlinear fractional programming, and Lagrange dual decomposition (Z. Zhou, Dong, Ota, & Chang, 2015). Matching theory provides a low-complexity decentralized self-organizing solution to the two-sided matching problem in college admissions (Gale & Shapley, 1962), marriage stability (Roth & Sotomayor, 1991), labor markets (O’Malley, 2007), etc., and has been widely applied for solving resource allocation problems in cellular networks (EI-Hajj, Dawy, & Saad, 2012; Pantiano, Bennis, Saad, & Debbah, 2015; Pantisano, Bennis, Saad, & Debbah, 2014a), cognitive radios (Feng et al., 2014), social networks (Gu, Zhang, Pan, & Han, 2015), D2D communications (Gu, Zhang, Pan, & Han, 2014), and mobile energy-harvesting networks (Niyato, Wang, Pink, Saad, & Kim, 2014), etc.