A Taxonomy and Survey of Energy-Efficient Data Centers and Cloud Computing Systems (1007.0066v2)

Abstract: Traditionally, the development of computing systems has been focused on performance improvements driven by the demand of applications from consumer, scientific and business domains. However, the ever increasing energy consumption of computing systems has started to limit further performance growth due to overwhelming electricity bills and carbon dioxide footprints. Therefore, the goal of the computer system design has been shifted to power and energy efficiency. To identify open challenges in the area and facilitate future advancements it is essential to synthesize and classify the research on power and energy-efficient design conducted to date. In this work we discuss causes and problems of high power / energy consumption, and present a taxonomy of energy-efficient design of computing systems covering the hardware, operating system, virtualization and data center levels. We survey various key works in the area and map them to our taxonomy to guide future design and development efforts. This chapter is concluded with a discussion of advancements identified in energy-efficient computing and our vision on future research directions.

A Taxonomy and Survey of Energy-Efficient Data Centers and Cloud Computing Systems

In the context of rising energy costs and environmental concerns, the paper "A Taxonomy and Survey of Energy-Efficient Data Centers and Cloud Computing Systems" by Beloglazov et al. offers a comprehensive review of the state-of-the-art research in energy-efficient computing. This involves critical analysis and categorization of existing techniques across several levels of computing infrastructure—hardware, operating system, virtualization, and data center levels. The main focus is on synthesizing and classifying the power and energy-efficient design approaches that have been proposed to date.

Causes and Problems of High Energy Consumption

The authors begin by addressing the fundamental issue: the increasing energy consumption driven by the proliferation of computing systems. This increase in energy use has been driven by growing demand from consumer, scientific, and business domains. They point out how the costs associated with energy use can often surpass the hardware costs over a server's lifetime. The problem extends to large-scale infrastructure like data centers, where cooling and power distribution become significant concerns.

Taxonomy of Power/Energy Management

The paper proposes a detailed taxonomy of power and energy management techniques:

• Hardware and Firmware Level: Involves strategies like Dynamic Component Deactivation (DCD) and Dynamic Performance Scaling (DPS), particularly focusing on methods such as Dynamic Voltage and Frequency Scaling (DVFS).
• Operating System Level: Covering real-time power management approaches like the Linux kernel's Ondemand governor, the ECOsystem framework, and the Nemesis OS, which act at various system resource levels.
• Virtualization Level: Discusses VM consolidation and power management at the hypervisor level using systems like Xen and VMware's DPM.
• Data Center Level: Strategies aimed at workload consolidation to minimize the number of active servers using approaches like Load Balancing and Unbalancing, and power-aware VM allocations.

Numerical Insights

The paper provides comprehensive insights into various numerical results of the discussed methodologies. For instance, the Ondemand governor in the Linux kernel dynamically adjusts CPU frequency based on usage, providing a trade-off between power savings and performance loss. Similarly, the GRACE project highlights that adapting software behavior based on power constraints can yield energy savings up to 32%. Data from empirical models, like those presented by Fan et al., demonstrate that power consumption grows almost linearly with CPU utilization, validating theoretical predictions.

Implications and Future Directions

The survey highlights several important implications. The insights help prioritize energy efficiency in the design and operation of data centers, recognizing trade-offs between performance and power consumption. For manufacturers and data center operators, the application of consolidated workloads and VM migrations offer substantial reductions in energy usage. In the field of Cloud computing, enhancing these techniques can lead to utility-based optimization, better QoS conformance, and potentially large-scale environmental benefits through minimized CO2 emissions.

The paper also speculates on future advancements:

• Emphasis on multi-core processors, which can involve more sophisticated energy-efficient strategies.
• Enhancing power management in network resources by improving the efficiency of networking hardware and optimizing data flows.
• Geographic workload distribution in Cloud federations to exploit varying energy costs and enhance cooling efficiency.
• Providing finer granularity of power management to end users by allowing application-specific optimizations.

Conclusion

Overall, this paper serves as an exhaustive guide for researchers and practitioners focused on developing energy-efficient techniques for data centers and Cloud computing. It makes a significant contribution by organizing and synthesizing existing efforts, identifying common characteristics, and suggesting potential future research paths. This work is invaluable in guiding the design and optimization of next-generation energy-efficient computing infrastructures.