Network auditing for low-cost assessment of networked equipment generated loads in office buildings

https://doi.org/10.1016/j.suscom.2013.04.002Get rights and content

Abstract

The large electricity expenditure of ubiquitous networked equipment within office buildings calls for dedicated load monitoring. Walk-throughs, in situ equipment-level power measurements, and installation of software agents on machines are generally conducted to assess generated loads. Such approaches have cost and time constraints that hinder widespread deployment in many sites if return on investment is not guaranteed. This work proposes to audit non-invasively equipment on virtual local area networks, enabling automated and low-cost assessment of networked equipment loads. Used as a preliminary load assessment, generated figures will be key to motivate further instrumentation. Experiments have been undertaken with the proposed methodology within a University Department comprising more than 850 unique networked machines. Results over a 15-month audit have highlighted that (1) client machines contributed to 18.4% of the Department's daytime power consumption units and to 14.6% of the nighttime power consumption units and (2) server machines contributed to 19.6% of the Department's daytime power consumption units and to 23% of the nighttime power consumption units; uncovering opportunities for important savings and fine-grained monitoring.

Introduction

With the waste of electricity consumed in buildings estimated at 30% [1], the service sector offers a fertile environment where opportunities for savings are significant. Examples of service sector employment include government, academia, banking and corporate offices. Users of related office buildings, the buildings’ electricity consumers, are not, in effect, the buildings’ owners, and are subsequently little concerned about energy spending. For instance, an office worker may not pay the same attention to switching off equipment at night as if he were at home, and would not take responsibility in switching off shared equipment such as printers, which may be needed by somebody else at some point. Such a situation typically results in important wastage [1] and little interest from building users in taking energy-saving actions [2]. It is therefore the responsibility of buildings owners to find ways to monitor equipment operation and reduce demand for energy.

Accounting of power consumption generally focuses on task-specific machines, plug loads, Lighting, Heating, Ventilation & Air Conditioning (L-HVAC) equipment, and networked equipment, especially in buildings of the service sector. The US Department of Energy [1], [2] has estimated the contribution of networked equipment to 20–30% in office buildings. Kawamoto et. al [3] also found that in Japan office equipment power management can save as much as 3.5 TWh per year.

Problems with after hours networked equipment power activity exist since the eve of computer infrastructures, despite the proliferation of monitoring tools and control systems. The vast majority of office buildings cannot, in effect, afford monitoring equipment and consultancy. Energy is not a core function or a major cost for businesses of the service sector, and financiers are generally unwilling to invest large amounts of money in energy efficiency improvements that could displace core capital investments, especially if return on investment is uncertain [15].

Energy awareness via general recommendations to users is, in many occasions, the only alternative. Building occupants are however not able to understand the impact of their actions without being provided with consumption and cost trend figures [4]. This work tackles the cost constraint associated to deploying electricity monitoring equipment, and proposes a novel technique based on network auditing and software data processing capable of assessing at low cost the networked equipment load demand within office buildings.

Section snippets

Related work

Automating the retrieval of networked equipment connection status on large-scale networks has driven various research and industrial efforts over the past decades, predominantly for uses over the Internet, e.g. Skitter's analysis of Internet topology and performance [5]. This work focuses on automating the retrieval of machines’ connection status on virtual local area networks (VLANs), in order to translate network activity reports into power activity reports. Capture of equipment power

Methodology

Assessing the power consumption of networked equipment from a central location does reduce costs and invasiveness, but introduces monitoring report inaccuracies.

Experimental setup

The choice of environment where to experiment the proposed load assessment technique was driven by discussions with the University College Dublin Building & Services department, which highlighted an important load demand at night and week-end for the building of the School of Computer Science & Informatics (SCSI). Access to the building's electrical measurement was made available. Similarly, information on the SCSI VLAN configuration was provided by the IT department and experimentation over

Future work and conclusions

Assessment of networked equipment load demand via VLAN auditing demonstrates that low-cost electricity proling can be achieved utilising existing building infrastructure. As a totally non-invasive approach, the accuracy of the load demand assessment is tightly linked to the accuracy of networked equipment power consumption estimations. The proposed approach estimates average power consumption figures at daytime and nighttime using both the typical distribution of networked equipment in office

Acknowledgements

The authors would like to thank Declan Delaney, for his help in analysing the experimental data. This work is supported by Science Foundation Ireland under grant 07/CE/I1147.

Anthony Schoofs has recently completed his PhD at CLARITY: Centre for Sensor Web Technologies in Ireland. His research has focused on innovative energy-efficient systems for large-scale buildings leveraging sensor systems, awarding him the Globe Sustainability Research award in 2011. He is currently CTO of Wattics Ltd, and spent several years working for Philips Research, in Eindhoven, the Netherlands.

References (31)

  • Powerman PC Manager,...
  • Electric Power Research Institute, nonintrusive appliance load monitoring system (nialms)—beta-test results, tr-108419,...
  • G. Hart

    Non-intrusive appliance load monitoring

    IEEE

    (1992)
  • A. Ruzzelli, G.O’Hare, A. Schoofs, C. Nicolas, Real-time recognition and profiling of appliances through a single...
  • Electric Power Research Institute, low-cost nialms technology—market issues & product assessment, tr-108918-v1, final...
  • Cited by (0)

    Anthony Schoofs has recently completed his PhD at CLARITY: Centre for Sensor Web Technologies in Ireland. His research has focused on innovative energy-efficient systems for large-scale buildings leveraging sensor systems, awarding him the Globe Sustainability Research award in 2011. He is currently CTO of Wattics Ltd, and spent several years working for Philips Research, in Eindhoven, the Netherlands.

    Gregory O'Hare completed his studies at the University of Ulster graduating with a B.Sc, M.Sc and Ph.D. He held the position of Head of the Department of Computer Science at University College Dublin (UCD) 2001–2004. Prior to joining UCD he has been on the Faculty of the University of Central Lancashire (1984–86) and the University of Manchester (1986–1996). He is an Associate Professor within the School of Computer Science & Informatics at UCD. His research interests are in the areas of Distributed Artificial Intelligence and Multi-Agent Systems (MAS), and Mobile & Ubiquitous Computing, Autonomic Systems and Wireless Sensor Networks.

    Antonio Ruzzelli is a Research Fellow team leader at CLARITY: Centre for Sensor technologies, located at University College Dublin, Ireland, and CEO of Wattics Ltd, an awarded start-up for electrical equipment-level monitoring and benchmarking. He is interested in low-power sensor systems with an emphasis on energy-efficiency in buildings, carbon footprinting and appliance recognition. Over the past years, he has served as principal investigator on a number of European/National projects in the area of sensor networking and intelligent buildings space.

    View full text