Proactive thermal-aware management in cloud datacenters.

摘要

The complexity of modern datacenters is growing at an alarming rate due to the rising popularity of the cloud-computing paradigm as an effective means to cater to the ever increasing demand for computing and storage. The management of modern datacenters is rapidly exceeding human ability, making autonomic approaches essential. In the meanwhile, the increasing demand for faster computing and high storage capacity has resulted in an increase in energy consumption and heat generation in datacenters. Due to the increased heat generation, cooling requirements have become a critical concern, both in terms of growing operating costs as well as their environmental and societal impacts. (e.g., increase in CO2 emissions, overloading the electric supply grid resulting in power cuts, heavy water usage for cooling systems causing water scarcity).;In this thesis, proactive thermal-aware datacenter management solutions, which include thermal- and energy-aware resource provisioning, cooling system optimization, and anomaly detection, are proposed to help minimize both the impact on the environment and the Total Cost of Ownership (TCO) of datacenters, making them energy efficient and green. For the proactive thermal-aware solutions, a novel architecture endowed with different abstract components is introduced, which is composed of four layers: the environment layer (which detects, localizes, characterizes, and tracks thermal hotspots), the physical-resource layer (which manages the hardware and software components of servers), the virtualization layer (which instantiates, configures, and manages VMs), and the application layer (which is aware of the workload's and applications' characteristics and behavior).;Our solutions autonomically manage datacenters using cross-layer information collected from the four-layered architecture and make decisions based on various application-specific optimization goals (e.g., performance, energy efficiency, anomaly detection rate). A sensing infrastructure to measure the datacenter's environmental change and methods to acquire thermal awareness (using real-time measurements and heat- and air-circulation models) are discussed. Then, specific proactive thermal-, energy-, and anomaly-aware solutions are proposed, which i) optimize cooling systems (i.e., air conditioner compressor duty cycle and fan speed) to prevent heat imbalance and minimize the cost of cooling, ii) maximize computing resource utilization to minimize datacenter energy consumption, and iii) differentiate servers' thermal map (temperature) frequently to maximize the thermal anomaly detection rate.