This specification defines media types for representing the future price of energy in JSON. It al... more This specification defines media types for representing the future price of energy in JSON. It also defines a way for a client device, such as an electric car, to discover a web server that can provide the future pricing for local electrical energy. This will allow the client device to make intelligent decisions about when to use energy. This draft is a very early skeleton of a draft to start discussion around this idea.
Networked end-systems such as desktops and set-top boxes are often left powered-on, but idle, lea... more Networked end-systems such as desktops and set-top boxes are often left powered-on, but idle, leading to wasted energy consumption. An alternative would be for these idle systems to enter low-power sleep modes. Unfortunately, today, a sleeping system sees degraded functionality: first, a sleeping device loses its network "presence" which is problematic to users and applications that expect to maintain access to a remote machine and, second, sleeping can prevent running tasks scheduled during times of low utilization (e.g., network backups). Various solutions to these problems have been proposed over the years including wake-on-lan (WoL) mechanisms that wake hosts when specific packets arrive, and the use of a proxy that handles idle-time traffic on behalf of a sleeping host. As of yet, however, an in-depth evaluation of the potential for energy savings, and the effectiveness of proposed solutions has not been carried out. To remedy this, in this paper, we collect data directly from 250 enterprise users on their end-host machines capturing network traffic patterns and user presence indicators. With this data, we answer several questions: what is the potential value of proxying or using magic packets? which protocols and applications require proxying? how comprehensive does proxying need to be for energy benefits to be compelling? and so on. We find that, although there is indeed much potential for energy savings, trivial approaches are not effective. We also find that achieving substantial savings requires a careful consideration of the tradeoffs between the proxy complexity and the idle-time functionality available to users, and that these tradeoffs vary with user environment. Based on our findings, we propose and evaluate a proxy architecture that exposes a minimal set of APIs to support different forms of idle-time behavior.
Most energy end uses such as space conditioning or water heating are apparently welldefined in wh... more Most energy end uses such as space conditioning or water heating are apparently welldefined in what is included, and have terminology that derives from the professionals who work in the relevant field. The topic of "miscellaneous" consumption lacks such clarity for historical and practical reasons. As this end use grows in size and interest for the energy community, the confusion and ambiguity around the topic is an increasing barrier to progress. This paper provides definitions for key terms and concepts with the intent that that future work can be more correctly and consistently reported and interpreted. In addition, it provides a taxonomy of product types and categories, which covers both residential and commercial miscellaneous consumption. A key element is identification of "electronics" as a distinct energy end use. Finally, products are identified as to whether they commonly have a low-power mode, and product types that have such modes within the traditional end uses are also listed.
The current ENERGY STAR specifications for imaging equipment address electricity consumption only... more The current ENERGY STAR specifications for imaging equipment address electricity consumption only in low-power modes. As products increase in functionality and speed, Active mode contributes to a greater portion of total product energy use. Additionally, consumer dissatisfaction with long recovery times can result in decreased use of energy-saving features. When EPA began to revise the imaging equipment specifications, the need to address Active mode for some products quickly became apparent. This was accomplished with the “typical electricity consumption” (TEC) approach that considers the electricity consumed by imaging equipment during its entire duty cycle. This method for assessing product energy efficiency has been received favorably by many stakeholders and demonstrates a forward-thinking approach to the development of an energy-efficiency specification. This paper reviews the process of creating the TEC method and resulting test procedure and specification levels. The authors...
Billions of dollars of electricity are being used to keep idle or unused network hosts fully powe... more Billions of dollars of electricity are being used to keep idle or unused network hosts fully powered-on only to maintain their network presence. We investigate how a Network Connectivity Proxy (NCP) could enable significant energy savings by allowing idle hosts to enter a low-power sleep state and still maintain full network presence. An NCP must handle ARP, ICMP, DHCP, and other low-level network presence tasks for a network host. An NCP must also be able to maintain TCP connections and UDP data flows and to respond to application messages. The focus of this paper is on how TCP connections can be kept alive during periods of host sleep by using a SOCKS-based approach called green SOCKS (gSOCKS) as part of an NCP. The gSOCKS includes awareness of the power state of a host. A prototype implementation of gSOCKS in a Linksys router shows that TCP connections can be preserved.
used to detect abnormal operations, and to conduct measurement and verification of energy or dema... more used to detect abnormal operations, and to conduct measurement and verification of energy or demand savings. This Agent was applied to analyze the energy use patterns in energy use patterns in a building at LBNL (one of the eight above mentioned sites) and a building at PNNL. In the LBNL building the change from normal operation to heating season operation was easily detected, and a particular feature of the load shape was investigated and determined to be caused by inefficient controls settings. The PNNL building exhibited load behavior that may indicate the potential for substantial energy savings. Researchers at Google investigated the model used in the M&V agent, comparing its results to those generated by other models, with a specific interest in predicting the peak daily load. As of the end of January, 2015, they reported that the LBNL model was the best performer by a small margin. In a larger study, the LBNL model was one of 10 models (both proprietary, and published) that were tested to evaluate their accuracy in predicting energy, specifically for M&V applications (Granderson et. al 2015 a; Granderson et al. 2015 b). The LBNL model performed well, demonstrating solid accuracy relative to the current state of the art. M&V with proxy occupancy data Occupancy data are typically not available for most buildings. Most common occupancy sensors used for lighting controls are not communicating sensors. They are commonly used in conference rooms, office areas, and rest rooms, but the data are not archived. Such data were not available for any of our test buildings, but "virtual sensing" (see above) provides a solution: data on the number of Wi-Fi connections as a function of time can serve as a proxy for occupancy data. We extended the statistical M&V model described above to add the number of Wi-Fi connections as a predictive variable. For two of the LBNL buildings with the virtual sensing we used the model with input data from November and December 2014 and January 2015, excluding the holiday period from December 23-January 1. We used the model predict the electric load as a function of time during the holiday period that was excluded from the input data. The model substantially outperformed the previous model: in both buildings, the use of the occupancy proxy variable reduced. The root-mean-squared error(RMSE) changed from 3.2 to 1.7 for one building, and from 21.7 to 17.5 in the other. These are 46 and 20% improvements respectively. We also compiled load data and Wi-Fi data from five UC Berkeley buildings, spanning four months in summer 2015. Four of the buildings showed little occupancy-related variation in load (and, indeed, little variation for any reason). In the fifth building, the occupancy proxy variable substantially improved the model predictions. In total, occupancy data substantially improved baseline electric load prediction in three of the seven test buildings. The other buildings showed little variation in load from day to day and week to week, in spite of moderate changes in occupancy patterns. Since the energy in most buildings is used to provide services to occupants, such as lighting and space conditioning, in an efficiently operated building the load would vary substantially with occupancy. The new capability to identify buildings whose load is insensitive to the number of occupants may allow those buildings to be targeted for energy savings measures such as occupancy-controlled lighting and ventilation. iii Conclusions and Future Directions While both of the occupancy-related aspects of this work (virtual occupancy sensing, and use of that virtual data to improve baseline model accuracy) are preliminary, they seem promising and merit further work in virtual sensing (also known as "proxy" or "inferential" sensing, and in the literature as "implicit sensing") and monitoring to generate an equivalent value to traditional sensors and yield a value that is actionable for building operations. We have identified promising future directions for each topic area. Among these will be to validate the IT system data with ground truth occupancy data, to compare the various methods with each other, and extract key metrics. Further work is also needed to use these and related occupancy data to better understand the linkages between occupancy and energy use for measurement and verification, fault diagnostics, energy forecasting, and improved building control.
etworks are one of the most significant developments in computing and a hallmark of modern societ... more etworks are one of the most significant developments in computing and a hallmark of modern society. However, along with increasing efficiency and productivity, both at home and in the workplace, networks have costs. One cost is the additional energy that electronic devices consume when attached to networks. Recent studies indicate that the economic and environmental impact of such energy use is substantial (www. eere.energy.gov/buildings/documents/ pdfs/office_telecom-vol1_final.pdf). In 1999, for example, office and network equipment-primarily PCs connected to the Internet-accounted for about 2 percent of US electricity consumption (http://enduse.lbl.gov/Info/LBNL-45917b.pdf). This roughly corresponds to $6 billion in energy costs and the emission of more than 50 million metric tons of carbon dioxide per year. During the next two decades, energy consumption by digitally networked devices is expected to increase at a faster rate than other types of energy use in buildings. Many such devices will not be PCs but emerging technologies designed for the home rather than the office.
Mo-va-on • Solar PV genera-on, baGery storage, and most loads are na-vely DC • How much efficienc... more Mo-va-on • Solar PV genera-on, baGery storage, and most loads are na-vely DC • How much efficiency savings with DC building distribu-on? • Par-cularly relevant for Zero Net Energy (ZNE) and microgrid buildings
Network connectivity has become nearly ubiquitous, and the energy use of the equipment required f... more Network connectivity has become nearly ubiquitous, and the energy use of the equipment required for this connectivity is growing. Network equipment consists of devices that primarily switch and route Internet Protocol (IP) packets from a source to a destination, and this category specifically excludes edge devices like PCs, servers and other sources and sinks of IP traffic. This paper presents the results of a study of network equipment energy use and includes case studies of networks in a campus, a medium commercial building, and a typical home. The total energy use of network equipment is the product of the stock of equipment in use, the power of each device, and their usage patterns. This information was gathered from market research reports, broadband market penetration studies, field metering, and interviews with network administrators and service providers. We estimate that network equipment in the USA used 18 TWh, or about 1% of building electricity, in 2008 and that consumption is expected to grow at roughly 6% per year to 23 TWh in 2012; world usage in 2008 was 51 TWh. This study shows that office building network switches and residential equipment are the two largest categories of energy use consuming 40% and 30% of the total respectively. We estimate potential energy savings for different scenarios using forecasts of equipment stock and energy use, and savings estimates range from 20% to 50% based on full market penetration of efficient technologies.
Storage, memory, processor, and communications bandwidth are all relatively plentiful and inexpen... more Storage, memory, processor, and communications bandwidth are all relatively plentiful and inexpensive. However, a growing expense in the operation of computer networks is electricity usage. Estimates place devices connected to the Internet as consuming about 2%, and growing, of the total electricity produced in the USA-much of this power consumption is unnecessary. Power management is needed to reduce this large and growing energy consumption of the Internet. We see power management as the 'next frontier' in research in computer networks. In this paper, we propose methods for reducing energy consumption of networked desktop computers. Using traffic characterization of university dormitory computers, we show that there is significant idle time that can be exploited for power management. However, current Ethernet adapters in desktop computers lack the capabilities needed to allow existing system power management features to be enabled. We address this problem with a proxying Ethernet adapter that handles routine network tasks for a desktop computer when it is in a low-power sleep mode. This proxying adapter can allow existing power management features in desktop computers to remain enabled and have the computer be 'on the network' at all times. The energy that we expect can be saved is in the range of 0.8-2.7 billion US dollars/year.
The IT equipment comprising the Internet in the USA uses about $6 billion of electricity every ye... more The IT equipment comprising the Internet in the USA uses about $6 billion of electricity every year. Much of this electricity use is wasted on idle, but fully powered-up, desktop PCs and network links. We show how to recover a large portion of the wasted electricity with improved power management methods that are focused on network issues.
This specification defines media types for representing the future price of energy in JSON. It al... more This specification defines media types for representing the future price of energy in JSON. It also defines a way for a client device, such as an electric car, to discover a web server that can provide the future pricing for local electrical energy. This will allow the client device to make intelligent decisions about when to use energy. This draft is a very early skeleton of a draft to start discussion around this idea.
Networked end-systems such as desktops and set-top boxes are often left powered-on, but idle, lea... more Networked end-systems such as desktops and set-top boxes are often left powered-on, but idle, leading to wasted energy consumption. An alternative would be for these idle systems to enter low-power sleep modes. Unfortunately, today, a sleeping system sees degraded functionality: first, a sleeping device loses its network "presence" which is problematic to users and applications that expect to maintain access to a remote machine and, second, sleeping can prevent running tasks scheduled during times of low utilization (e.g., network backups). Various solutions to these problems have been proposed over the years including wake-on-lan (WoL) mechanisms that wake hosts when specific packets arrive, and the use of a proxy that handles idle-time traffic on behalf of a sleeping host. As of yet, however, an in-depth evaluation of the potential for energy savings, and the effectiveness of proposed solutions has not been carried out. To remedy this, in this paper, we collect data directly from 250 enterprise users on their end-host machines capturing network traffic patterns and user presence indicators. With this data, we answer several questions: what is the potential value of proxying or using magic packets? which protocols and applications require proxying? how comprehensive does proxying need to be for energy benefits to be compelling? and so on. We find that, although there is indeed much potential for energy savings, trivial approaches are not effective. We also find that achieving substantial savings requires a careful consideration of the tradeoffs between the proxy complexity and the idle-time functionality available to users, and that these tradeoffs vary with user environment. Based on our findings, we propose and evaluate a proxy architecture that exposes a minimal set of APIs to support different forms of idle-time behavior.
Most energy end uses such as space conditioning or water heating are apparently welldefined in wh... more Most energy end uses such as space conditioning or water heating are apparently welldefined in what is included, and have terminology that derives from the professionals who work in the relevant field. The topic of "miscellaneous" consumption lacks such clarity for historical and practical reasons. As this end use grows in size and interest for the energy community, the confusion and ambiguity around the topic is an increasing barrier to progress. This paper provides definitions for key terms and concepts with the intent that that future work can be more correctly and consistently reported and interpreted. In addition, it provides a taxonomy of product types and categories, which covers both residential and commercial miscellaneous consumption. A key element is identification of "electronics" as a distinct energy end use. Finally, products are identified as to whether they commonly have a low-power mode, and product types that have such modes within the traditional end uses are also listed.
The current ENERGY STAR specifications for imaging equipment address electricity consumption only... more The current ENERGY STAR specifications for imaging equipment address electricity consumption only in low-power modes. As products increase in functionality and speed, Active mode contributes to a greater portion of total product energy use. Additionally, consumer dissatisfaction with long recovery times can result in decreased use of energy-saving features. When EPA began to revise the imaging equipment specifications, the need to address Active mode for some products quickly became apparent. This was accomplished with the “typical electricity consumption” (TEC) approach that considers the electricity consumed by imaging equipment during its entire duty cycle. This method for assessing product energy efficiency has been received favorably by many stakeholders and demonstrates a forward-thinking approach to the development of an energy-efficiency specification. This paper reviews the process of creating the TEC method and resulting test procedure and specification levels. The authors...
Billions of dollars of electricity are being used to keep idle or unused network hosts fully powe... more Billions of dollars of electricity are being used to keep idle or unused network hosts fully powered-on only to maintain their network presence. We investigate how a Network Connectivity Proxy (NCP) could enable significant energy savings by allowing idle hosts to enter a low-power sleep state and still maintain full network presence. An NCP must handle ARP, ICMP, DHCP, and other low-level network presence tasks for a network host. An NCP must also be able to maintain TCP connections and UDP data flows and to respond to application messages. The focus of this paper is on how TCP connections can be kept alive during periods of host sleep by using a SOCKS-based approach called green SOCKS (gSOCKS) as part of an NCP. The gSOCKS includes awareness of the power state of a host. A prototype implementation of gSOCKS in a Linksys router shows that TCP connections can be preserved.
used to detect abnormal operations, and to conduct measurement and verification of energy or dema... more used to detect abnormal operations, and to conduct measurement and verification of energy or demand savings. This Agent was applied to analyze the energy use patterns in energy use patterns in a building at LBNL (one of the eight above mentioned sites) and a building at PNNL. In the LBNL building the change from normal operation to heating season operation was easily detected, and a particular feature of the load shape was investigated and determined to be caused by inefficient controls settings. The PNNL building exhibited load behavior that may indicate the potential for substantial energy savings. Researchers at Google investigated the model used in the M&V agent, comparing its results to those generated by other models, with a specific interest in predicting the peak daily load. As of the end of January, 2015, they reported that the LBNL model was the best performer by a small margin. In a larger study, the LBNL model was one of 10 models (both proprietary, and published) that were tested to evaluate their accuracy in predicting energy, specifically for M&V applications (Granderson et. al 2015 a; Granderson et al. 2015 b). The LBNL model performed well, demonstrating solid accuracy relative to the current state of the art. M&V with proxy occupancy data Occupancy data are typically not available for most buildings. Most common occupancy sensors used for lighting controls are not communicating sensors. They are commonly used in conference rooms, office areas, and rest rooms, but the data are not archived. Such data were not available for any of our test buildings, but "virtual sensing" (see above) provides a solution: data on the number of Wi-Fi connections as a function of time can serve as a proxy for occupancy data. We extended the statistical M&V model described above to add the number of Wi-Fi connections as a predictive variable. For two of the LBNL buildings with the virtual sensing we used the model with input data from November and December 2014 and January 2015, excluding the holiday period from December 23-January 1. We used the model predict the electric load as a function of time during the holiday period that was excluded from the input data. The model substantially outperformed the previous model: in both buildings, the use of the occupancy proxy variable reduced. The root-mean-squared error(RMSE) changed from 3.2 to 1.7 for one building, and from 21.7 to 17.5 in the other. These are 46 and 20% improvements respectively. We also compiled load data and Wi-Fi data from five UC Berkeley buildings, spanning four months in summer 2015. Four of the buildings showed little occupancy-related variation in load (and, indeed, little variation for any reason). In the fifth building, the occupancy proxy variable substantially improved the model predictions. In total, occupancy data substantially improved baseline electric load prediction in three of the seven test buildings. The other buildings showed little variation in load from day to day and week to week, in spite of moderate changes in occupancy patterns. Since the energy in most buildings is used to provide services to occupants, such as lighting and space conditioning, in an efficiently operated building the load would vary substantially with occupancy. The new capability to identify buildings whose load is insensitive to the number of occupants may allow those buildings to be targeted for energy savings measures such as occupancy-controlled lighting and ventilation. iii Conclusions and Future Directions While both of the occupancy-related aspects of this work (virtual occupancy sensing, and use of that virtual data to improve baseline model accuracy) are preliminary, they seem promising and merit further work in virtual sensing (also known as "proxy" or "inferential" sensing, and in the literature as "implicit sensing") and monitoring to generate an equivalent value to traditional sensors and yield a value that is actionable for building operations. We have identified promising future directions for each topic area. Among these will be to validate the IT system data with ground truth occupancy data, to compare the various methods with each other, and extract key metrics. Further work is also needed to use these and related occupancy data to better understand the linkages between occupancy and energy use for measurement and verification, fault diagnostics, energy forecasting, and improved building control.
etworks are one of the most significant developments in computing and a hallmark of modern societ... more etworks are one of the most significant developments in computing and a hallmark of modern society. However, along with increasing efficiency and productivity, both at home and in the workplace, networks have costs. One cost is the additional energy that electronic devices consume when attached to networks. Recent studies indicate that the economic and environmental impact of such energy use is substantial (www. eere.energy.gov/buildings/documents/ pdfs/office_telecom-vol1_final.pdf). In 1999, for example, office and network equipment-primarily PCs connected to the Internet-accounted for about 2 percent of US electricity consumption (http://enduse.lbl.gov/Info/LBNL-45917b.pdf). This roughly corresponds to $6 billion in energy costs and the emission of more than 50 million metric tons of carbon dioxide per year. During the next two decades, energy consumption by digitally networked devices is expected to increase at a faster rate than other types of energy use in buildings. Many such devices will not be PCs but emerging technologies designed for the home rather than the office.
Mo-va-on • Solar PV genera-on, baGery storage, and most loads are na-vely DC • How much efficienc... more Mo-va-on • Solar PV genera-on, baGery storage, and most loads are na-vely DC • How much efficiency savings with DC building distribu-on? • Par-cularly relevant for Zero Net Energy (ZNE) and microgrid buildings
Network connectivity has become nearly ubiquitous, and the energy use of the equipment required f... more Network connectivity has become nearly ubiquitous, and the energy use of the equipment required for this connectivity is growing. Network equipment consists of devices that primarily switch and route Internet Protocol (IP) packets from a source to a destination, and this category specifically excludes edge devices like PCs, servers and other sources and sinks of IP traffic. This paper presents the results of a study of network equipment energy use and includes case studies of networks in a campus, a medium commercial building, and a typical home. The total energy use of network equipment is the product of the stock of equipment in use, the power of each device, and their usage patterns. This information was gathered from market research reports, broadband market penetration studies, field metering, and interviews with network administrators and service providers. We estimate that network equipment in the USA used 18 TWh, or about 1% of building electricity, in 2008 and that consumption is expected to grow at roughly 6% per year to 23 TWh in 2012; world usage in 2008 was 51 TWh. This study shows that office building network switches and residential equipment are the two largest categories of energy use consuming 40% and 30% of the total respectively. We estimate potential energy savings for different scenarios using forecasts of equipment stock and energy use, and savings estimates range from 20% to 50% based on full market penetration of efficient technologies.
Storage, memory, processor, and communications bandwidth are all relatively plentiful and inexpen... more Storage, memory, processor, and communications bandwidth are all relatively plentiful and inexpensive. However, a growing expense in the operation of computer networks is electricity usage. Estimates place devices connected to the Internet as consuming about 2%, and growing, of the total electricity produced in the USA-much of this power consumption is unnecessary. Power management is needed to reduce this large and growing energy consumption of the Internet. We see power management as the 'next frontier' in research in computer networks. In this paper, we propose methods for reducing energy consumption of networked desktop computers. Using traffic characterization of university dormitory computers, we show that there is significant idle time that can be exploited for power management. However, current Ethernet adapters in desktop computers lack the capabilities needed to allow existing system power management features to be enabled. We address this problem with a proxying Ethernet adapter that handles routine network tasks for a desktop computer when it is in a low-power sleep mode. This proxying adapter can allow existing power management features in desktop computers to remain enabled and have the computer be 'on the network' at all times. The energy that we expect can be saved is in the range of 0.8-2.7 billion US dollars/year.
The IT equipment comprising the Internet in the USA uses about $6 billion of electricity every ye... more The IT equipment comprising the Internet in the USA uses about $6 billion of electricity every year. Much of this electricity use is wasted on idle, but fully powered-up, desktop PCs and network links. We show how to recover a large portion of the wasted electricity with improved power management methods that are focused on network issues.
Uploads
Papers by Bruce Nordman