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

Abstract:

Large-scale cloud infrastructure resource management requires improving data center energy efficiency while guaranteeing Quality of Service (QoS) and detecting server performance variability caused by hardware or software failures. Dynamic Virtual Machine (VM) consolidation works mostly address energy efficiency, but they lack comprehensive, scalable, and low-overhead approaches that address energy efficiency and performance variability.

They also assume over-simplistic power models and fail to account for all the delay and power costs of VM migration and host power mode transition. In modern heterogeneous servers, these assumptions lead to unrealistic or inefficient results.

This paper proposes a centralized-distributed low-overhead failure-aware dynamic VM consolidation strategy to reduce energy consumption in large data centers. A distributed multi-agent Machine Learning (ML) strategy selects the best power mode and frequency for each host during runtime, and a centralized heuristic migrates the VMs.

Our Multi-AGent machine learNing-based approach for Energy efficienT dynamIc Consolidation (MAGNETIC) is implemented in a modified CloudSim simulator and considers the energy and delay overheads associated with host power mode transition and VM migration.

It is evaluated using power traces from various workloads running in real servers and resource utilization logs from cloud data center infrastructures. Our strategy reduces data center energy consumption by 15% compared to the state-of-the-art (SoA), guarantees QoS, and reduces VM migrations and host power mode transitions by 86 and 90%, respectively. It also has the best scalability, requiring less than 0.7% time overhead for a data center with 1,500 VMs. Finally, our solution automatically migrates VMs from failing hosts and drains them of workload when host performance variability is detected.

Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.