Models and Metrics to Enable Energy-Efficiency Optimizations

Abstract

P u b l i s h e d b y t h e I E E E C o m p u t e r S o c i e t y• circuit techniques such as disabling the clock signal to a processor's unused parts; • architectural techniques such as replacing complex uniprocessors with multiple simple cores; and • support for multiple low-power states in processors, memory, and disks.At the system level, the latter approach requires policies to intelligently exploit these low-power states for energy savings. Across multiple systems in a cluster or data center, these policies can involve dynamically adapting workload placement or power provisioning to meet specific energy or thermal goals. To facilitate these optimizations, we need metrics to define energy efficiency, which will help designers compare designs and identify promising energy-efficient technologies. We also need models to predict the effects of dynamic power management policies, particularly over many systems. Unlike the significant body of work on power management and optimization, there has been relatively little focus on metrics and models.We address the challenges in defining metrics for energy efficiency with a specific case study on JouleSort, which provides a complete, full-system benchmark for energy efficiency across a variety of system classes. Power consumption and energy efficiency are important factors in the initial design and day-to-day management of computer systems. Researchers and system designers need benchmarks that characterize energy efficiency to evaluate systems and identify promising new technologies.To predict the effects of new designs and configurations, they also need accurate methods of modeling power consumption.

Energy Efficiency Effects of Vectorization in Data Reuse Transformations for Many-Core Processors—A Case Study †

“…These counters can be accessed either by the RAPL (Running Average Power Limit ) interface (root-level) or the powercap interface (user-level). We report both core/package-energy consumption and Energy Delay Product (EDP: Joule × Second) [30,31] to perform energy efficiency analysis. For both, lower values corresponds to better energy efficiency.…”

Seeking for the Optimal Energy Modelisation Accuracy to Allow Efficient Datacenter Optimizations

“…Moreover, it allows comparing power and performance among different servers and serves as a toolset for use in improving server efficiency. Nevertheless, these SPEC energy models, only take into account the CPU utilization to determine the energy consumption of the whole system (as it is the resource that consumes most of the host's power [9]). This can lead to some approximations or errors.…”

“…This can lead to some approximations or errors. Actually, we know that disk and network operations, for instance, have an impact on the total server consumption [9]. In order to get more accurate energy prediction for the compute nodes, other metrics should be taken into account and not only the CPU.…”

Review of performance metrics for green data centers: a taxonomy study

“…Typically data center benchmarking is carried out under a certain workload, which is artificially generated or produced in real life: real or artificial workload → benchmark → performance metrics. Although some benchmark workloads are available, for example, server-side Java under various loads for SPECpower [23], external sort for JouleSort [32,33], data center benchmarking is generally taken under normal data center practice and workload.…”

“…There are two major and complementary methods [33] to build a "green" data center: (1) involve "green" elements in the design and building process of a data center, (2) "greenify" the process of running and operating a data center in everyday usage. Various research activities have been carried out to manage data centers in a "green" mode (the latter method), such as reducing data center temperature [28,39,42], increasing server utilization [21,26,37], and decreasing power consumption of computing resources [7,8,12,20].…”