Abstract:Energy consumption constraints on computing systems are more important than ever. Maintenance costs for high performance systems are limiting the applicability of processing devices with large dissipation power. New solutions are needed to increase both the computation capability and the power efficiency. Moreover, energy efficient applications should balance performance vs. consumption. Therefore power data of components are important. This work presents the most remarkable alternatives to measure the power consumption of different types of computing systems, describing the advantages and limitations of available power measurement systems. Finally, a methodology is proposed to select the right power consumption measurement system taking into account precision of the measure, scalability and controllability of the acquisition system.Keywords: power measurement; energy consumption profiling; energy efficiency; instrumentation; power analysis
Many researchers have analyzed how to retrieve values of EC from computing systems and make use of them in EEAs. Nevertheless, this issue is not standardized and there are different methods to retrieve the EC. In this section, relevant works related to power measurement and optimization solutions for EEAs are presented, describing the advantages and drawbacks of each system.
The main advantage of this alternative it is the facility to configure and start measurements. Although the scalability of the solution is possible, once the measurement system is setup the flexibility is low due to limitations of internal built-in circuit configuration. Besides, the accuracy could not fit the EEA needs, and the acquisition frequency may be too low for a correct EEA profiling. This type of MD may be of interest for quick tests and temporal analysis of EEAs; however the cost of IPS equipment which is usually expensive is a disadvantage.
Other external MD alternative could be the setup of a specific measurement system according to the EEA needs, for example: non-invasive, accuracy, cost, etc. making available some separate measures of the computing architecture. This specific hardware system could be deployed introducing additional physical sensors located in the input power lines of any potential DUT of the computing architecture. These solutions can provide a wide range of alternatives, such as the usage of current sensors based on hall-effect providing a non-intrusive solution (see Figure 2(b)).
Thus, it is proposed a methodology to select the right power consumption measurement system taking into account precision of the measure, scalability and controllability of the acquisition system in order to manage the performance-consumption ratio efficiently. The measurement systems based on hardware sensors can be developed to measure the EC according to the end user needs. However the new processing solutions formed by several computing units integrated in a single die, difficult the measure process of internal components due to inaccessibility of separated power lines. Therefore, considering the integrated power models the best approach for new fused architectures, which can help programmers to profile their code and design EEAs.
The MSR approach to power measurement represents an innovative solution for EC profiling. Although this feature nowadays is hardware vendor dependent, the initiative to introduce energy counters will be standardized in the near future of computing systems. This approach will turn feasible to retrieve power consumption from complex computing systems with different processing units, network interfaces, memories, etc. at software level, enabling the development of power aware systems.
In Budget 2016, the Government of Canada announced its commitment to work with stakeholders to develop a performance measurement framework for business accelerators and incubators (BAIs) in Canada. The first step toward increased collaboration among BAIs in Canada to create a national solution for data collection and performance reporting took place on February 10, 2017 in Toronto, Ontario. Leaders from 18 BAIs discussed opportunities and challenges of creating a national performance measurement framework.
A BAI Steering Committee consisting of a representative group of BAI leaders and policy makers was subsequently formed to continue an inclusive national discussion and provide leadership in crafting a national performance measurement solution that works for the BAI community and its partners in government. The Committee's overarching mandate has been to work alongside the Government of Canada in partnership to a) establish a performance measurement framework, and b) pilot a performance measurement platform for BAIs. The primary output of the Committee's work was a Performance Measurement Framework (PMF) launched in February 2018.
In addition to presenting an updated survey instrument with clear definitions for the key performance metrics, this document serves as an operating manual for the BAI performance measurement process. As such, part II of the report provides the necessary background for onboarding new BAI participants and government partners, including the initial rationale for establishing a national performance measurement framework and a simple logic model that guides the design of the PMF. The report describes the approach for collecting, analyzing and reporting the data, including the methodology that will be used by approved researchers to produce the descriptive statistics and econometric analyses that will illuminate the relationship between BAI programs and the economic performance of client firms. Finally, the report details the operations and administration of the performance measurement platform, including processes for obtaining consent to share information and protecting the confidentiality of data.
This report is intended to guide the BAI community and its partners in government as they proceed with the next phase of building a national performance measurement solution. It presents an updated Performance Measurement Framework (termed BAI PMF 2.0) which will form the basis of a national performance measurement solution, documents the progress achieved during the pilot process, and reflects the key decisions and design inputs of the pilot participants and public sector partners. Specifically, the report: 2b1af7f3a8