This paper focuses on the power management of state-of-the-art networked devices (like common PCs, servers, set-top boxes, etc.) to evaluate their behavior, model their internal dynamics and possible sources of inefficiency, and optimize their performance and energy efficiency. To this purpose, we started from an experimental characterization of the power management schemes in common device platforms based on commercial off-the-shelf hardware and open-source software (i.e., common PCs/servers devices running the Linux operating system). The characterization allowed us to formalize an analytical model able to accurately capture the power management dynamics at hardware (at ACPI level and beyond) and software levels (Linux Governors). Finally, the proposed model has been applied to analyze the efficiency of networked devices according to various configurations of internal parameters and incoming workload. Thanks to its intrinsic accuracy and the representation of different fine-grained details, the model is able to provide precious information on the possible sources of inefficiency, and on how to act on policy parameters to optimize the system behavior. © 2014 Elsevier B.V. All rights reserved.
Modeling power management in networked devices
Bruschi R.;Lago P.;
2014-01-01
Abstract
This paper focuses on the power management of state-of-the-art networked devices (like common PCs, servers, set-top boxes, etc.) to evaluate their behavior, model their internal dynamics and possible sources of inefficiency, and optimize their performance and energy efficiency. To this purpose, we started from an experimental characterization of the power management schemes in common device platforms based on commercial off-the-shelf hardware and open-source software (i.e., common PCs/servers devices running the Linux operating system). The characterization allowed us to formalize an analytical model able to accurately capture the power management dynamics at hardware (at ACPI level and beyond) and software levels (Linux Governors). Finally, the proposed model has been applied to analyze the efficiency of networked devices according to various configurations of internal parameters and incoming workload. Thanks to its intrinsic accuracy and the representation of different fine-grained details, the model is able to provide precious information on the possible sources of inefficiency, and on how to act on policy parameters to optimize the system behavior. © 2014 Elsevier B.V. All rights reserved.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.