Research Projects

Below are brief summaries to highlight current ongoing projects, along with selected project publications and funding sources. This page only summarizes a fraction of our work. For more information, please see the publications page.

smartDesigning Smart* Homes

The Smart* project focuses on new opportunities for optimizing and scheduling the energy consumption of modern “smart homes,” which include pervasive energy sensing and actuation capabilities. As part of the project, we have heavily instrumented several real homes to serve as experimental testbeds for our research. We have publicly released much of our data and software tools from our deployments. In general, our work adapts resource management techniques from operating systems and networking to manage resources and energy in buildings.

panelModel-based Building Energy Data Analytics

A significant barrier to improving building energy efficiency is that fine-grained monitoring of electrical loads at large scales remains impractical: it is expensive, invasive, and unreliable. The project’s goal is to advance our understanding of building energy consumption using a model-based paradigm to design a wide range of novel non-intrusive energy data data analytics using a building’s smart meter data. Our analytics combine domain-specific knowledge of power systems with sophisticated machine learning techniques to reveal new insights into building energy usage.

Supported by NSF grant #1253063

datacenterSystem Support for Transient Servers

Today’s distributed applications are built under the implicit assumption that the underlying servers will be mostly available, aside from occasional faults. However, in many emerging scenarios, data centers and clouds now provide transient availability, such that servers are available temporarily for an uncertain amount of time. Transience in modern distributed systems arises in many contexts, including green data centers powered using renewable energy sources, cloud platforms that sell servers in dynamic spot markets, and smart data centers that curtail their energy usage based on signals from the smart grid. This project focuses on designing distributed systems that treat transience as a first-class design objective. In particular, since traditional fault-tolerance methods are too expensive or not well suited to handling the intermittent availability of transient servers, we focus on low-cost transience mechanisms for modern data center and cloud platforms.

Supported by NSF #1422245

lightSmart Grid Security and Privacy

While improved networked sensors and data analytics are enabling a new level of visibility into building activity and energy usage, they also raise new security and privacy concerns. To prevent these concerns from discouraging broader adoption of “smart” technologies, it is important to ensure buildings are able to control the information they leak to third parties, including utilities, cloud platforms, etc. Our research focuses on low-cost and non-intrusive methods for ensuring consumer security and privacy in buildings that include pervasive networked sensing and actuation.

solarDesigning Net-Zero Data Centers and Clouds

Despite energy-efficiency improvements, data center energy demands continue to rise, increasing worldwide by an estimated 56% from 2005-2010 and accounting for an estimated 1.7-2.2% of U.S. electricity usage. This project focuses on designing net-zero data centers that operate, either directly or indirectly, off intermittent renewable energy sources. Thus, our focus is primarily on designing data center applications to operate efficiently in the presence of significant and frequent power fluctuations from renewable energy sources. Our project benefits from the close proximity of the new Massachusetts Green High Performance Computing Center (MGHPCC) in Holyoke, Massachusetts.

Supported by NSF grant #1339839