ETAS: Energy and thermal‐aware dynamic virtual machine consolidation in cloud data center with proactive hotspot mitigation

Abstract

Data centers consume an enormous amount of energy to meet the ever-increasing demand for cloud resources. Computing and Cooling are the two main subsystems that largely contribute to energy consumption in a data center. Dynamic Virtual Machine (VM) consolidation is a widely adopted technique to reduce the energy consumption of computing systems. However, aggressive consolidation leads to the creation of local hotspots that has adverse effects on energy consumption and reliability of the system. These issues can be addressed through efficient and thermal-aware consolidation methods. We propose an Energy and Thermal-Aware Scheduling (ETAS) algorithm that dynamically consolidates VMs to minimize the overall energy consumption while proactively preventing hotspots. ETAS is designed to address the trade-off between time and the cost savings and it can be tuned based on the requirement. We perform extensive experiments by using the real-world traces with precise power and thermal models.The experimental results and empirical studies demonstrate that ETAS outperforms other state-of-the-art algorithms by reducing overall energy without any hotspot creation. KEYWORDScloud computing, data center cooling, energy efficiency in a data center, hotspots VM consolidation INTRODUCTIONCloud computing is a massive paradigm shift from how the computing capabilities are acquired in past from traditional ownership model to current subscription model. 1 Cloud offers on-demand access to elastic resources as services with pay as you go model based on the actual usage of resources. Cloud data centers are the backbone infrastructure to cloud services. To adapt to the increasing demand for massive scale cloud services, data centers house thousands of servers to fulfill their computing needs. However, they are power hungry and consume a huge amount of energy to provide cloud services in a reliable manner. According to the USA energy department report, 2 data centers in the USA itself consume about 2% (70 billion kWh) of the total energy production. Not only do data centers consume huge power; they significantly contribute to the greenhouse gas emissions resulting in high carbon footprints. To be precise, they generate 43 million tons of CO 2 per year and continues to grow at an annual rate of 11%. 3 If the necessary steps are taken, data center power consumption can be reduced from the predicted worst case of 8000 TWh to 1200 TWh by the year 2030. 4 Therefore, improving the energy efficiency of the cloud data center is quintessential for sustainable and cost-effective cloud computing.A significant part of cloud data centers' energy consumption emanates from computing and cooling systems. In particular, the contribution of cooling system power is almost equal to the computing system. 5 In this context, a data center resource management system should holistically contemplate computing and cooling power together to achieve overall energy efficiency.In pursuance of reducing the computing energy, workloads are consolidated on the fewest hosts as ...

iThermoFog: IoT‐Fog based automatic thermal profile creation for cloud data centers using artificial intelligence techniques

“…19 Integration of CloudSim, ThermoSim and iFogsim is described in Figure 2. We use 12 cloud VMs and 18 fog devices in our setup, same as done in ROUTER application by Gill et al 10 As ThermoSim and CloudSim have been widely validated in previous work, 3,4,10,11 we expect that these results could be reproduced in realistic settings.…”

“…This work however does not consider other types of utilization and energy consumption profiles arising from disks, network, and memory, which are critical to consider in Fog computing. Ilager et al 4 propose an energy and thermal-aware scheduling algorithm (ETAS) that dynamically consolidates VMs to minimize the overall energy consumption while proactively preventing thermal hot spots. Another limitation is that the algorithm assumes a static cooling environment, which may not be versatile to different cooling settings.…”

“…For feasible solutions, we consider a truncated GMM, ignoring the negative samples. To generate the training data, we run various pre-existing scheduling policies like random allocation, ETAS, 4 and others 12,14 to form a dataset of temperature values for different scheduling task and host parameters, and schedules. Using this dataset, we find * and Σ * , which we then deploy to predict the temperature of the all possible scheduling decisions (provided by CloudSim) at the current state, given by the iFogSim simulator, and take the one with least ∥ ∥.…”

“…3 Currently, temperature management techniques use reactive mechanisms to generate temperature profiles and control the cooling facilities available within the CDC. 4 But, as shown in prior work, 5 such mechanisms can have large delays in profiling the thermal characteristics of the CDC, which rapidly increase the energy consumption and operational costs. 6 To solve this problem, a novel approach is required that is quickly able to adapt to dynamic task characteristics and host resource utilization in a heterogeneous environment that can not only keep temperature under control but also reduce energy, latency and bandwidth usage.…”

CloudSimSDN‐NFV: Modeling and simulation of network function virtualization and service function chaining in edge computing environments

“…Louis et al extended the CloudSim to simulate the storage power consumption in the data centers. Ilager et al also extended the components of the server energy model in CloudSim by taking runtime temperature into consideration. They modeled the hot spots generated by the running physical hosts, which have a significant impact on the energy cost for the whole cooling and power solution of the cloud and edge computing.…”