An energy-aware combinatorial auction-based virtual machine scheduling model and heuristics for green cloud computing
- Resource Type
- Authors
- ÖZER, ALİ HAYDAR
- Source
- Subject
- General Computer Science
Sinyal İşleme
Mühendislik
ENGINEERING
Information Systems, Communication and Control Engineering
Cloud computing
Bilgisayar Bilimleri
Electrical and Electronic Engineering
Engineering, Computing & Technology (ENG)
Genel Bilgisayar Bilimi
ENGINEERING, ELECTRICAL & ELECTRONIC
Resource scheduling
Virtual machine placement
Computer Sciences
Elektrik ve Elektronik Mühendisliği
Mühendislik, Bilişim ve Teknoloji (ENG)
COMPUTER SCIENCE
Energy-aware
Combinatorial auctions
Fizik Bilimleri
Genetic algorithm
Signal Processing
Physical Sciences
Engineering and Technology
Bilgisayar Bilimi
MÜHENDİSLİK, ELEKTRİK VE ELEKTRONİK
Mühendislik ve Teknoloji
Bilgi Sistemleri, Haberleşme ve Kontrol Mühendisliği
- Language
- English
Considering the increasing demand for cloud computing, and the financial and environmental impact of the increasing energy consumption trend of data centers, improving energy efficiency is vital for cloud service providers. In this study, an energy-aware virtual machine scheduling model is proposed which is based on the multi-unit nondiscriminatory combinatorial auction. The model includes a powerful bidding language that allows users to declare their complicated virtual machine requests using logical AND and OR relations along with the time constraints. The study also presents the formal definition of the model and the associated optimization problem for determining the optimum schedule and energy-efficient placement of VMs on physical servers. The optimization problem is formulated using integer linear programming and several heuristic solution methods including the Genetic Algorithm are proposed for this problem. The performances of the model and the proposed heuristics are assessed on a comprehensive test suite. The proposed model is estimated to provide approximately a 37% improvement in revenues, and the solution methods are estimated to provide high-quality solutions within only 5% of the optimum which enable the model to be deployed in large-scale clouds.