Part 3. The Case for Energy-Proportional Computing

Wide dynamic power range
Current desktop and server processors can consume less than one-third of their peak power at very-low activity modes, creating a dynamic range of more than 70 percent of peak power. CPUs targeted at the mobile or embedded markets can do even better, with idle power often reaching one-tenth or less of peak power. 10 They achieve this even when not using any performance-impacting—or software-visible—energy-saving modes.

In our experience, the dynamic power range of all other components is much narrower: less than 50 percent for DRAM, 25 percent for disk drives, and 15 percent for networking switches.

Active low-power modes
A processor running at a lower voltage-frequency mode can still execute instructions without requiring a performance-impacting mode transition. It is still active. There are no other components in the system with active low-power modes. Networking equipment rarely offers any low-power modes, and the only low-power modes currently available in mainstream DRAM and disks are fully inactive. That is, using the device requires paying a latency and energy penalty for an inactive-to-active mode transition. Such penalties can significantly degrade the performance of systems idle only at submillisecond time scales.


Figure 4. Power usage and energy efficiency in a more energy-proportional server. This server has a power efficiency of more than 80 percent of its peak value for utilizations of 30 percent and above, with efficiency remaining above 50 percent for utilization levels as low as 10 percent.

Compared to today's machines, servers with a dynamic power range of 90 percent, shown in Figure 4, could cut by one-half the energy used in data center operations. 5 They would also lower peak power at the facility level by more than 30 percent, based on simulations of real-world data center workloads. These are dramatic improvements, especially considering that they arise from optimizations that leave peak server power unchanged. The power efficiency curve in Figure 4 fundamentally explains these gains. This server has a power efficiency of more than 80 percent of its peak value for utilizations of 30 percent and above, with efficiency remaining above 50 percent for utilization levels as low as 10 percent.

In addition to its energy-savings potential, energy-proportional hardware could obviate the need for power management software, or at least simplify it substantially, reducing power management to managing utilization.

Fundamentally, the latency and energy penalties incurred to transition to the active state when starting an operation make an inactive energy-savings mode less useful for servers. For example, a disk drive in a spun-down, deep-sleep state might use almost no energy, but a transition to active mode incurs a latency penalty 1,000 times higher than a regular access latency. Spinning up the platters also carriers a large energy penalty. Such a huge activation penalty restricts spin-down modes to situations in which the device will be idle for several minutes; this rarely occurs in servers. On the other hand, inactive energy-savings modes with wake-up penalties of only a small fraction of the regular operations' latency are more likely to benefit the server space, even if their low-energy state operates at relatively higher energy levels than would be possible in deep-sleep modes.

Active energy-savings schemes, by contrast, are useful even when the latency and energy penalties to transition to a high-performance mode are significant. Since active modes are operational, systems can remain in low-energy states for as long as they remain below certain load thresholds. Given that periods of low activity are more common and longer than periods of full idleness, the overheads of transitioning between active energy-savings modes amortize more effectively.

Servers and desktop computers benefit from much of the energy-efficiency research and development that was initially driven by mobile devices' needs. However, unlike mobile devices, which idle for long periods, servers spend most of their time at moderate utilizations of 10 to 50 percent and exhibit poor efficiency at these levels. Energy-proportional computers would enable large additional energy savings, potentially doubling the efficiency of a typical server. Some CPUs already exhibit reasonably energy-proportional profiles, but most other server components do not.

We need significant improvements in memory and disk subsystems, as these components are responsible for an increasing fraction of the system energy usage. Developers should make better energy proportionality a primary design objective for future components and systems. To this end, we urge energy-efficiency benchmark developers to report measurements at nonpeak activity levels for a more complete characterization of a system's energy behavior.

Acknowledgments

We thank Xiaobo Fan and Wolf-Dietrich Weber for coauthoring the power provisioning study that motivated this work, and Catherine Warner for her comments on the manuscript.

References

US Environmental Protection Agency, "Report to Congress on Server and Data Center Energy Efficiency: Public Law 109-431"; www.energystar.gov/ia/partners/prod_development/downloads/EPA_Datacenter_Report_Congress_Final1.pdf.
J.G. Koomey, "Estimating Total Power Consumption by Servers in the U.S. and the World"; http://enterprise.amd.com/Downloads/svrpwrusecompletefinal.pdf.
L.A. Barroso, "The Price of Performance: An Economic Case for Chip Multiprocessing," ACM Queue, Sept. 2005, pp. 48-53.
J. Hill et al., "System Architecture Directions for Networked Sensors," Proc. SIGOPS Oper. Syst. Rev., ACM Press, vol. 34, no. 5, 2000, pp. 93-104.
X. Fan, W.-D. Weber, and L.A. Barroso, "Power Provisioning for a Warehouse-Sized Computer"; http://research.google.com/archive/power_provisioning.pdf.
S. Ghemawat, H. Gobioff and S.-T. Leung, "The Google File System"; www.cs.rochester.edu/meetings/sosp2003/papers/p125-ghemawat.pdf.
S. Siddha, V. Pallipadi, and A. Van De Ven, "Getting Maximum Mileage Out of Tickless," Proc. 2007 Linux Symp., 2007, pp. 201-208.
E. Ravussin et al., "Determinants of 24-Hour Energy Expenditure in Man: Methods and Results Using a Respiratory Chamber"; http://www.pubmedcentral.nih.gov/picrender.fcgi?artid=423919&blobtype=pdf.
E.F. Coyle, "Improved Muscular Efficiency Displayed as Tour de France Champion Matures"; http://jap.physiology.org/cgi/reprint/98/6/2191.
Z. Chen et al., "A 25W(max) SoC with Dual 2GHz Power Cores and Integrated Memory and I/O Subsystems"; www.pasemi.com/downloads/PA_Semi_ISSCC_2007.pdf.
Source: computer