A Case for Energy-Proportional Computing
📜 Abstract
Redesigning computing systems to be energy proportional could bring large energy savings across the information and communication technology infrastructure, increasing its sustainability. Energy efficiency at both level of individual systems and the entire facility is an increasing concern for large computing installations. We observe that energy-proportional computers will be able to provide the most meaningful gains as the fraction of consumption that is unavoidable (proportional to peak performance) drops. An energy-proportional design requires significant improvements in the energy usage profile of every system component, particularly the memory and disk subsystems.
✨ Summary
This paper presents a compelling argument for designing computing systems that are energy-proportional as a means of achieving significant energy savings across IT infrastructures. Energy-proportional computing involves designing systems where the energy used is proportional to the system’s workload. The authors, Luis Andre Barroso and Urs Hölzle, elaborate that the most significant impacts of such designs would be seen when energy consumption that is unavoidable (related to peak system performance) is reduced. They highlight the need for advancements in all system components, notably memory and disk subsystems, to approach true energy-proportional designs.
Energy-proportional computing has had a pronounced influence on how modern data centers are conceptualized and structured. By emphasizing the correlation between workload and energy use, this approach has prompted further research and development in enhancing hardware efficiency, leading to advancements in processor power management and component-level energy control. Though the paper itself does not cite specific subsequent works, the concept of energy-proportional computing is widely recognized in subsequent research on green computing and eco-friendly IT infrastructure development.