Full length article
Profit maximization in cache-aided intelligent computing networks

https://doi.org/10.1016/j.phycom.2023.102065Get rights and content

Abstract

This paper focuses on the profit maximization problem in a reconfigurable intelligent surfaces (RIS) aided computing network, where multiple heterogeneous users offload their computational tasks to one computational access point (CAP) for seeking computing acceleration at the cost of profit. In particular, the CAP can also pre-store a part of the computing task to speed up computing, and the system has limited communication and computing resources, where heterogeneous users have different offloading requirements and the CAP can dynamically allocate the system resources to meet the requirements of users to earn profits. To maximize the system profit, we devise the system by proposing a resource allocation scheme which employs a genetic algorithm (GA), based on statistical channel state information (CSI) of wireless links. The proposed algorithm maximizes the long-term profit of the system by optimizing resource allocation among users. Finally, simulation results are provided to verify the proposed scheme. The results show that our proposed resource allocation scheme outperforms the conventional ones.

Introduction

In recent years, the rapid development of internet technology and the widespread use of mobile devices have led to an exponential increase in the amount of data that needs to be transmitted over wireless networks [1], [2], [3], [4]. However, due to the long transmission distance between mobile devices and the cloud data center, the traditional centralized computing model requires all data to be migrated to the cloud center for processing, resulting in a significant transmission delay. This calculation mode will seriously reduce the service quality of mobile devices and damage the user’s service experience, and even cannot satisfy the application requirement of high real-time demand. To solve this problem, researchers have proposed mobile edge computing (MEC) in recent years [5], [6], [7], [8], [9].

In the MEC architecture, edge computing service providers offer cloud storage and cloud computing services for mobile devices. [10], [11], [12]. The computing power provided by the cloud service provider can create a carrier-class service environment with high performance, low latency, and high bandwidth. Mobile users can achieve ultra-low computing latency by offloading computing tasks of applications to the edge server for computation. Some scholars propose to offload a part of the task to the cloud for computing, which can effectively reduce the delay in transmission and the cost of computing in the cloud. In [13], [14], they focus on the smart internet of vehicle (IoV) based on multi-user MEC and use a linear combination of latency and energy consumption to analyze system performance and derive optimal offloading ratios by minimizing system cost. In [15], [16], the authors examine a multi-user MEC-assisted IoV by integrating budget constraints into the system design. They propose a deep reinforcement learning (DRL) approach combined with a convex optimization algorithm.

In wireless communication networks, the randomness of wireless channels and the severe fading of channels can lead to weak energy and information received by the receiver. In recent years, reconfigurable intelligent surfaces (RIS), as a new type of device, have garnered significant attention from scholars [17]. RIS is a new communication method for Beyond 5G (B5G) networks, which is a planar array consisting of a large number of reconfigurable passive components. The intelligent controller can independently control each element to produce a specific amount of phase shift on the incident signal, which can modify the propagation of the reflected signal. RIS consist of a large number of low-cost passive reflective elements that can be intelligently controlled through signals to enhance the performance of wireless communication networks. In [12], the authors investigate the outage performance of MEC networks and propose two different RIS unit selection criteria. Additionally, based on these two RIS selection criteria, the authors derive corresponding exact analytical expressions and asymptotic expressions, respectively, to gain insights into the MEC network. In [7], the authors study RIS-assisted MEC networks under physical layer security and propose a secure data transmission model. The authors employ deep deterministic policy gradient (DDPG) to optimize system performance.

The development of edge computing has given birth to the “edge service market” computing model. In the side service market, computing resources are regarded as a kind of “commodity”, and users and server providers are in a “buy-sell” relationship [18], [19], [20], [21]. Mobile users do not have to care who provides the service, and the provider finds the target customer through the price [22], [23], [24]. However, in the case of multiple users and service providers, the user task scheduling constrained computing resource allocation and suppliers. Therefore, the user and supplier’s pricing strategy cannot simply be modeled as a linear constraint problem to solve in the edge computing services market. In order to explore the reward mechanism for maximizing the profit of resource providers in edge computing, in [25] the authors propose an incentive mechanism in a non-competitive environment that uses a profit-maximizing model to explore how edge clouds charge mobile devices. Some scholars have studied the profit maximization problem in collaborative computing networks and proposed a migratory bird optimization method based on a simulated annealing algorithm to solve the profit maximization problem [24], [26]. Overall, IRS technology reduces latency and cost by improving communication links. However, as far as we know, there still needs to be more work to optimize the long-term benefits of users in edge networks, which further promotes the work of this paper.

This paper focuses on the revenue optimization problem of MEC servers with limited spectrum resources, and A CSI-based GA method is proposed to maximize the long-term benefit of CAP. The main contributions are as follows:

(1) This paper proposes a long-term profit optimization model for the reconfigurable intelligent surface assisted computing network to focus on maximizing profit.

(2) We obtain the closed-form solution for the system outage probability in the RIS-assisted unloading scenario.

(3) To obtain the optimal bandwidth allocation strategy, we propose a GA method based on statistical CSI.

Section snippets

System model

As depicted in Fig. 1, we consider an RIS assisted MEC network, where there are N users denoted by {Sn|1nN} can transmit task to the single antenna CAP to compute via an RIS. The RIS consists of N zones, where each zone is equipped with one passive reflection element. Specially, each zone has a different phase shift, so we assume that the reflection unit in each region can guide the signal to a specific user. In addition, the CAP is also equipped with a cache space, which can pre-cache a

System optimization

In the second chapter, we propose two important indicators: a time delay and energy consumption during task unloading. However, when providing services, the edge cloud will consume a lot of wireless frequency resources and computing resources, so how to maximize the profit of edge cloud providers is also an important issue. To this end, we have defined a new performance index, namely CPA profit Φ, and considered the costs of CAP in the cache, task transmission and calculation, respectively.

In

Numerical and simulation results

In this section, we present some numerical and simulation results to validate our analysis. Without loss of generality, we set αn=0.5 βn=0.5, the system total communication bandwidth B is set to 5MHz, the transmission power and calculation power are set as Pt=1.2W and Pc=3W, respectively. In further, the task of users is uniformly distributed in the range of [2,15] Mb. The channels in the network follow Rayleigh flat fading [27], [28], [29]. Moreover, the CPU-cycles frequency at the CAP is set

Conclusions

This work studied the profit maximization problem in RIS-aided computing networks, where CAP can pre-cache and compute user-offloaded tasks. We consider practical communication scenarios with limited bandwidth resources, and due to the uneven distribution of spectrum resources, it is impossible to maximize the CAP profit. We propose a genetic algorithm method based on statistical channel state information (CSI) to optimize bandwidth resources for the considered system. In particular, the method

CRediT authorship contribution statement

Rui Zhao: Conceptualization, Methodology, Software, Investigation, Formal analysis, Writing – original draft. Fusheng Zhu: Software, Validation. Maobing Tang: Data curation, Writing – original draft, Visualization, Investigation. Le He: Conceptualization, Funding acquisition, Resources, Supervision, Writing – review & editing, Software, Validation.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This work was supported by This work was supported by the Key-Area Research and Development Program of Guangdong Province, China (no. 2019B090904014), the Science and Technology Program of Guangzhou (No. 202201010047), in part by the National Natural Science Foundation of China under Grant 61977018.

Rui zhao received the B.E. degree in computer science and technology from Bohai University, Jinzhou, China, in 2018, and the master’s degree from the computer technology, Guangzhou University, Guangzhou, China, in 2021. He is currently working with the School of Computer Science, Guangzhou University. His current research interests include machine learning and mobile edge computing resource scheduling algorithms.

References (29)

  • WuW. et al.

    Unmanned aerial vehicle swarm-enabled edge computing: Potentials, promising technologies, and challenges

    IEEE Wirel. Commun.

    (2022)
  • ZhouW. et al.

    Priority-aware resource scheduling for UAV-mounted mobile edge computing networks

    IEEE Trans. Vehic. Tech.

    (2023)
  • CaiD. et al.

    On the performance of NOMA with hybrid ARQ

    IEEE Trans. Veh. Technol.

    (2018)
  • ZhouW. et al.

    Profit maximization for cache-enabled vehicular mobile edge computing networks

    IEEE Trans. Vehic. Tech.

    (2023)
  • Cited by (0)

    Rui zhao received the B.E. degree in computer science and technology from Bohai University, Jinzhou, China, in 2018, and the master’s degree from the computer technology, Guangzhou University, Guangzhou, China, in 2021. He is currently working with the School of Computer Science, Guangzhou University. His current research interests include machine learning and mobile edge computing resource scheduling algorithms.

    Fusheng Zhu graduated from Huazhong University of Science and Technology in 1996, receiving his B.E. in electronic and information engineering, and received his M.B.A. degree from Fudan University in 2011. His research direction mainly includes the 6G mobile network and B5G vertical application, and network communication.

    Maobin Tang is currently an Associate Professor with the School of Computer Science and Educational Software, Guangzhou University. His research directions include big data and artificial intelligence.

    Le He received the bachelor’s degree from Northeastern University in 2018. He is currently pursuing the master’s degree with the Network and Intelligent Computing Technology (NICT) Laboratory, School of Computer Science, Guangzhou University. His current research interests include machine learning, signal processing, and their real world applications.

    View full text