Abstract:We address the problem of under-utilization of resources in datacenters during cloud operations, specifically focusing on the challenge of online virtual machine (VM) scheduling. Rather than following the traditional approach of scheduling VMs based solely on their static flavors, we take into account their dynamic CPU utilization. We employ Gamma -robustness theory to manage the dynamic nature and introduce a novel variant of bin packing -(), which theoretically protects the Physical Machines (PMs) from hotspots formation within a specified probability alpha . We develop a scheduling algroithm named CloseRadiusFit and cold-start AI-based prediction algorithms for the online version of . To verify the quality of our approach towards the optimal solutions, we solve the Offline problem by designing a novel Mixed Integer Linear Programming (MILP) model and a combination of numerical upper and lower bounds. Our experimental results demonstrate that CloseRadiusFit achieves narrow gaps of 1.6% and 3.1% when compared to the lower and upper bounds, respectively. Note to Practitioners -A growing trend in the cloud industry involves overcommitting VMs on PMs. While this approach can ease the problem of low utilization of resources in datacenters, it also introduces a higher risk of hotspots due to resource contention and competition among VMs. In this work, we propose a novel method that leverages Gamma -robustness theory and introduce effective heuristics to achieve ultimate utilization of datacenter resources while ensuring desirable service quality. We validate our approach using real-world production data from Huawei Cloud, improving resource utilization by 125% over traditional flavor-based allocation methods, while maintaining the occurrence of hotspots below 5% ( alpha=0.05 ). Our solution only requires VMs' real utilization data that is typically already collected in cloud providers' production environments. Therefore, with minimal modifications to the existing scheduling system, cloud providers can easily implement our solution and reap its benefits. Moreover, in cases of the absence of historical utilization data for VMs (cold-start), we use machine learning to predict VM utilization statistics for our approach.

Carbon-Aware Computing for Data Centers with Probabilistic Performance Guarantees

Carbon-Aware Computing for Datacenters

Carbon-Aware Computing in a Network of Data Centers: A Hierarchical Game-Theoretic Approach

Probability-Based Online Algorithm for Switch Operation of Energy Efficient Data Center

Carbon Responder: Coordinating Demand Response for the Datacenter Fleet

Using Geographic Load Shifting to Reduce Carbon Emissions

Cooling-Aware Energy and Workload Management in Data Centers Via Stochastic Optimization

Spatio-temporal load shifting for truly clean computing

Uncertainty-Aware Decarbonization for Datacenters

Proactive Demand Response for Data Centers: A Win-Win Solution

CarbonScaler: Leveraging Cloud Workload Elasticity for Optimizing Carbon-Efficiency

Quality Time: Carbon-Aware Quality Adaptation for Energy-Intensive Services

Cost-Effective Request Scheduling for Greening Cloud Data Centers

Online job scheduling scheme for low-carbon data center operation: An information and energy nexus perspective

On the Limitations of Carbon-Aware Temporal and Spatial Workload Shifting in the Cloud

Battery-Assisted Online Operation of Distributed Data Centers With Uncertain Workload and Electricity Prices

The Environmental Potential of Hyper-Scale Data Centers: Using Locational Marginal CO$_2$ Emissions to Guide Geographical Load Shifting

Robust Workload and Energy Management for Sustainable Data Centers

Hotspot-Aware Scheduling of Virtual Machines with Overcommitment for Ultimate Utilization in Cloud Datacenters

Use of time-series based forecasting technique for balancing load and reducing consumption of energy in a cloud data center

Greening Geo-distributed Data Centers by Joint Optimization of Request Routing and Virtual Machine Scheduling