MAGNETIC - Multi-Agent Machine Learning-Based Approach for Energy Efficient Dynamic Consolidation in Data Centers

Aug 5, 2026

Upload by Ophelia

Improving the energy efficiency of data centers while guaranteeing Quality of Service (QoS), and detecting performance variability caused by hardware or software failures, are key challenges in large-scale cloud infrastructure resource management. Prior dynamic VM consolidation works often target energy alone, rely on oversimplified power models, and neglect delay and power costs of VM migration and host power mode transitions. This paper proposes a low-overhead failure-aware strategy minimizing energy consumption via multi-agent ML runtime power-mode and frequency selection plus centralized VM migration. Implemented in CloudSim with real-server power traces and workload logs, it reduces energy by up to 15% while maintaining QoS, cutting VM migrations and power mode transitions by up to 86% and 90%, with under 0.7% time overhead, and auto-migrating VMs from failing hosts.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages14

Reading Time35 min

AccessFree to download

Introduction

Energy efficiency and QoS challenges in data centers

Dynamic VM consolidation and power mode transitions

Limitations of prior migration-aware approaches

Motivation and positioning of MAGNETIC

MAGNETIC: Multi-Agent Machine Learning-Based Approach for Energy Efﬁcient Dynamic Consolidation in Data Centers Kawsar Haghshenas, Ali Pahlevan, Student Member, IEEE , Marina Zapater, Member, IEEE, Siamak Mohammadi, senior Member, IEEE, and David Atienza, Fellow, IEEE Abstract—Improving the energy efﬁciency of data centers while guaranteeing Quality of Service (QoS), together with detecting performance variability of servers caused by either hardware or software failures, are two of the major challenges for efﬁcient resource management of large-scale cloud infrastructures. Previous works in the area of dynamic Virtual Machine (VM) consolidation are mostly focused on addressing the energy challenge, but fall short in proposing comprehensive, scalable, and low-overhead approaches that jointly tackle energy efﬁciency and performance variability. Moreover, they usually assume over-simplistic power models, and fail to accurately consider all the delay and power costs associated with VM migration and host power mode transition. These assumptions are no longer valid in modern servers executing heterogeneous workloads and lead to unrealistic or inefﬁcient results. In this paper, we propose a centralized-distributed low-overhead failure-aware dynamic VM consolidation strategy to minimize energy consumption in large-scale data centers. Our approach selects the most adequate power mode and frequency of each host during runtime using a distributed multi-agent Machine Learning (ML) based strategy, and migrates the VMs accordingly using a centralized heuristic. Our Multi-AGent machine learNing-based approach for Energy efﬁcienT dynamIc Consolidation (MAGNETIC) is implemented in a modiﬁed version of the CloudSim simulator, and considers the energy and delay overheads associated with host power mode transition and VM migration, and is evaluated using power traces collected from various workloads running in real servers and resource utilization logs from cloud data center infrastructures. Results show how our strategy reduces data center energy consumption by up to 15% compared to other works in the state-of-the-art (SoA), guaranteeing the same QoS and reducing the number of VM migrations and host power mode transitions by up to 86% and 90%, respectively. Moreover, it shows better scalability than all other approaches, taking less than 0.7% time overhead to execute for a data center with 1500 VMs. Finally, our solution is capable of detecting host performance variability due to failures, automatically migrating VMs from failing hosts and draining them from workload. Index Terms—Host Power Mode, Machine Learning, Migration Cost, Power Mode Transition Cost, VM Consolidation, VM Migration, Energy Efﬁciency, Cloud Data Centers. ~~ ~~ F ~~ ~~ 1 INTRODUCTION HIGH energy consumption and performance variability prob lems are two major challenges in modern cloud data centers, as they greatly affect operational expenses, total cost of ownership and revenue [1] . Global data center electricity usage accounted for 1.1-1.5% of total electricity use in 2010 [2] and increases at yearly rate of 2.1%[3] . As most data centers rely on fossil fuels as their main energy source, huge energy consumption also results in high carbon emissions and environmental concerns. However, according to a recent report by Shehabi et al. in 2016 [4], a potential of 45% reduction in electricity demand of data centers can be achieved compared to current trends, by improving their energy efﬁciency. Moreover, for the signiﬁcantly large data centers, hardware failures and software anomalies are more frequent, leading to application performance variability and, eventually, Quality of Service (QoS) degradation [5] . 􀀏 K. Haghshenas and S. Mohammadi are with the School of Electrical and Computer Engineering, University of Tehran, Tehran 17469-37181, Iran E-mail: fkhaghshenas, [smohamadi](smohamadig@ut.ac.ir)[g](smohamadig@ut.ac.ir)[@ut.ac.ir](smohamadig@ut.ac.ir) 􀀏 K. H

FAQs

What problem does MAGNETIC address in data center resource management?

MAGNETIC addresses reducing energy consumption in data centers while preserving QoS, and detecting performance variability caused by hardware or software failures.

How does the proposed approach decide host power behavior and VM placement?

It selects the most adequate power mode and frequency for each host at runtime using a distributed multi-agent ML strategy, then migrates VMs via a centralized heuristic.

What costs does MAGNETIC explicitly model when migrating VMs?

It considers energy and delay overheads of both host power mode transitions and VM migration, avoiding oversimplified power models used by prior work.

Aug 5, 2026

Upload by Ophelia