1 / 8

Energy-Efficient Computing and Computing for Efficient Energy Usage

Energy-Efficient Computing and Computing for Efficient Energy Usage. Yanlei Diao and Prashant Shenoy Department of Computer Science University of Massachusetts, Amherst. Energy-Efficient Computing Goal : make computation energy efficient Tools : Hardware, software techniques.

amil
Download Presentation

Energy-Efficient Computing and Computing for Efficient Energy Usage

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Energy-Efficient Computing and Computing for Efficient Energy Usage Yanlei Diao and Prashant Shenoy Department of Computer Science University of Massachusetts, Amherst Yanlei Diao, University of Massachusetts Amherst

  2. Energy-Efficient Computing Goal: make computation energy efficient Tools: Hardware, software techniques Two Aspects of Sustainable Computing • Computing for Efficient Energy Usage • Goal: green physical infrastructures and the natural environment • Tools: hardware, software, algorithms, optimizations Yanlei Diao, University of Massachusetts Amherst

  3. Monitoring Energy Usage in Buildings • Buildings consume 75% of the electricity in the US! • Sensor-driven energy monitoring a first step to greening of buildings • Sensors in office and residential buildings • Occupancy sensors: detect human presence and track movement • Outlet-level sensors: detect usage at individual outlets (e.g., AC plugs, tap) • Meter-level sensors: detect total usage of buildings (gas, water, electricity) • Our first deployment: instructed an 1700 sq. ft home • 35 outlets and all 33 wall switches monitored every 2 seconds • Meter-level sensor tracking usage every second for 6+ months • Second deployment: deployed 60 outlet sensors in the CS building Yanlei Diao, University of Massachusetts Amherst

  4. Sensor Deployment and Appliance Signatures Example: Refrigerator over a 24-hour period Yanlei Diao, University of Massachusetts Amherst

  5. Large-Scale Deployments in the Future • Large-scale deployments in the future: • All rooms/zones in a building • All buildings in a district • All buildings in a city A real-time view of energy consumption in a building/district/city. Help develop strategies and trigger actions for more efficient usage. Yanlei Diao, University of Massachusetts Amherst

  6. Data Analysis for Efficient Energy Usage • Consumer-view: real-time fine-grain usage streams • Utility view: smart meter streams from 100,000’s of homes • Energy bill capping / peak usage reduction • Use real-time usage data to cap/reduce total and peak consumption • Energy conservation • Identify zones with no active usage, turn off lights/HVAC systems • Correlate occupancy and usage sensor streams to detect these conditions • Anomaly detection: detect unusual usage patterns and alert users • Scaling issues • How to manage 100,000 or million streams from smart meters? • Large-scale, real-time data collection and data analysis is key • Infrastructure for data collection and dissemination • Real-time detection of patterns/trends; integrating usage with billing, identifying real-time incentives; comparing real-time usage with history • Parallel stream processing… Yanlei Diao, University of Massachusetts Amherst

  7. Energy-Efficient Data Storage Systems • Eenergy-efficient hardware: Flash memory, SSD’s • Software solutions for data management • Storage-centric sensor networks: • exploit energy-efficient flash storage on sensor nodes • reduce expensive communication • Large database systems: • use flash-based storage for both data and indexes • deal with expensive random writes in index design Yanlei Diao, University of Massachusetts Amherst

  8. Energy-Efficient DBMS: New Opportunities • A DBMS optimized for energy efficiency • Single machine: Limited opportunity for software-based energy optimization [SIGMOD’10] • The highest performing configuration is the most energy-efficient. • Shared disks / disk farms: computing and storage are decoupled • SSDs make the storage system energy proportional. • How do we make computing energy proportional? Consolidation? Sharing? • How do we make it scale? • Numerous embedded devices (in the smart planet context): flash memory and CPU integrated within a microcontroller • Flash memory requires the entire chip to operate at a higher voltage. • Lower voltage causes errors in flash writes. • Correct errors using software solutions? Yanlei Diao, University of Massachusetts Amherst

More Related