GMHA: Growable Meta-Heuristic Algorithm for Multi-Objective Optimization Problems and its Application in Cloud Scheduling

Scheduling problems in distributed computing (such as Cloud or edge computing) usually belong to multi-objective optimization problems (MOPs). Meta-heuristic algorithm (MHA) is an effective type of contemporary algorithm for solving difficult MOPs. The improvement of MHA is an urgent spot, whose mai...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:IEEE transactions on services computing s. 1 - 15
Hlavní autoři: Zhou, Guangyao, Xu, Minxian, Zhang, Jiashu, Buyya, Rajkumar
Médium: Journal Article
Jazyk:angličtina
Vydáno: IEEE 2025
Témata:
Benchmark testing
Cloud computing
Cloud Scheduling
Convergence
Energy consumption
Genetic algorithms
Growable Meta-Heuristic
Metaheuristics
Multi-Objective Optimization Problem
Particle swarm optimization
Processor scheduling
Schedules
Scheduling
ISSN:1939-1374, 2372-0204
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Scheduling problems in distributed computing (such as Cloud or edge computing) usually belong to multi-objective optimization problems (MOPs). Meta-heuristic algorithm (MHA) is an effective type of contemporary algorithm for solving difficult MOPs. The improvement of MHA is an urgent spot, whose main challenge lies in accelerating convergence speed while enhancing the optimality of convergence solutions. This challenge requires a universal and effective method to improve the search efficiency of various MHAs throughout the entire iteration process for solving MOPs. Towards this target, this paper restructures the MHAs' framework uniformly, and proposes a universal growable meta-heuristic algorithm framework (GMHA) with hybrid multi-growth routes. Providing the flexibility for the combination of various algorithms to serve as the growth route, GMHA is applicable to diverse MHAs. For the sake of the adaptability of GMHA for various MOPs including Cloud scheduling, the paper further establishes several general growth routes, including equidistant feasible solution search route (EFSS), gradient neighborhood search route (GNS), and weighted neighborhood search route (WNS). Statistical test on various MOP benchmarks demonstrate that GMHA has a probability of <inline-formula><tex-math notation="LaTeX">90.15\%</tex-math></inline-formula> to improve the performance of various MHAs. Compared to the corresponding MHA in experiments on multi-objective Cloud scheduling problems, GMHA achieves 2.116 times the average convergence speed, with the specific reductions of <inline-formula><tex-math notation="LaTeX">25.05\%</tex-math></inline-formula> in total energy consumption, <inline-formula><tex-math notation="LaTeX">4.94\%</tex-math></inline-formula> in the maximum energy consumption of server nodes, and <inline-formula><tex-math notation="LaTeX">35.12\%</tex-math></inline-formula> in the sum of the standard deviations of utilizations.
ISSN:1939-1374
2372-0204
DOI:10.1109/TSC.2025.3610714