An effective fitness dependent optimizer algorithm for edge server allocation in mobile computing
Walaa H. El-Ashmawi,Adam Slowik,Ahmed F. Ali
DOI: https://doi.org/10.1007/s00500-023-09582-y
IF: 3.732
2024-01-12
Soft Computing
Abstract:Due to the speedy development of mobile communication devices, the traditional cloudlet computing networks struggle to manipulate the huge collected data from these devices. Mobile computing has been developed to overcome the access delay (AD) and the workload balance (WB) in the traditional cloudlet networks by transferring the computing process from the remote servers in the cloudlet to the edge servers (ESs) allocated closer to the mobile users. The searching for the optimal place to allocate the ESs in mobile edge computing is the main challenge and it considers an NP-hard problem. In this work, we propose a new objective function with threefold objectives, the AD between base stations and ESs, the WB among various controlled ESs, and the energy consumption (EC) of ESs. Therefore, we formulate the edge server allocation problem as a multi-objective optimization problem that requires an efficient optimizer algorithm for solving it. To minimize the proposed objective function, we present an effective fitness-dependent optimizer (EFDO) algorithm and test it on Shanghai Telecom's BS dataset. To investigate the efficiency of the proposed EFDO algorithm, we compare it against seven algorithms taken from literature (e.g., K-means, Random, TopFirst, PSO, CSA, Jaya, and JS). The numerical results verified the superiority of the proposed algorithm in terms fitness function reached up to 14.84%, 18.56%, 17.28%, and 13.48% when compared with PSO, CSA, Jaya, and JS, respectively with various number of base stations and edge servers. Although k-means has been achieved the least average AD among compared algorithms which considered only the distance between the BS and ES, the proposed algorithm has achieved the superior results in WB ranged from 12 to 72% when compared against other algorithms. In addition, it has gained the least average EC among the compared algorithms.
computer science, artificial intelligence, interdisciplinary applications