CoPPer: Soft Real-Time Application Performance Using Hardware Power Capping

CoPPer: Soft Real-Time Application Performance Using Hardware Power Capping. Imes, C., Zhang, H., Zhao, K., & Hoffmann, H. In 2019 IEEE International Conference on Autonomic Computing (ICAC), pages 31-41, June, 2019.
doi abstract bibtex 1 download

Dynamic voltage and frequency scaling (DVFS) has been the cornerstone of innumerable software approaches to meeting application timing requirements with minimal energy. However, recent trends in technology-e.g., moving voltage converters on chip-favor hardware control of DVFS, as hardware can both react faster to external events and perform fine-grained power management across a device. We respond to these trends with CoPPer, which instead uses hardware power capping to meet application performance requirements with high energy efficiency. We find that meeting performance requirements with power capping is more challenging than using DVFS because the relationship between power and performance is non-linear and has diminishing returns at high power values. CoPPer overcomes these difficulties by using adaptive control to approximate non-linearities and a novel gain limit to avoid over-allocating power when it is no longer beneficial. We evaluate CoPPer with 20 parallel applications and compare it to both a classic linear DVFS controller and to a sophisticated control-theoretic, model-driven software DVFS manager. CoPPer provides all the functionality of the sophisticated DVFS-based approach, without requiring a user-specified model or time-consuming, exhaustive application/system pre-characterization. Compared to DVFS, CoPPer's gain limit reduces energy by 6% on average and by 12% for memory-bound applications. For high performance requirements, the energy savings are even greater: 8% on average and 18% for memory-bound applications.

@INPROCEEDINGS{8831193,
  author={Imes, Connor and Zhang, Huazhe and Zhao, Kevin and Hoffmann, Henry},
  booktitle={2019 IEEE International Conference on Autonomic Computing (ICAC)},
  title={CoPPer: Soft Real-Time Application Performance Using Hardware Power Capping},
  year={2019},
  volume={},
  number={},
  pages={31-41},
  abstract={Dynamic voltage and frequency scaling (DVFS) has been the cornerstone of innumerable software approaches to meeting application timing requirements with minimal energy. However, recent trends in technology-e.g., moving voltage converters on chip-favor hardware control of DVFS, as hardware can both react faster to external events and perform fine-grained power management across a device. We respond to these trends with CoPPer, which instead uses hardware power capping to meet application performance requirements with high energy efficiency. We find that meeting performance requirements with power capping is more challenging than using DVFS because the relationship between power and performance is non-linear and has diminishing returns at high power values. CoPPer overcomes these difficulties by using adaptive control to approximate non-linearities and a novel gain limit to avoid over-allocating power when it is no longer beneficial. We evaluate CoPPer with 20 parallel applications and compare it to both a classic linear DVFS controller and to a sophisticated control-theoretic, model-driven software DVFS manager. CoPPer provides all the functionality of the sophisticated DVFS-based approach, without requiring a user-specified model or time-consuming, exhaustive application/system pre-characterization. Compared to DVFS, CoPPer's gain limit reduces energy by 6% on average and by 12% for memory-bound applications. For high performance requirements, the energy savings are even greater: 8% on average and 18% for memory-bound applications.},
  keywords={energy conservation;multiprocessing systems;power aware computing;soft real-time application performance;hardware power capping;application timing requirements;minimal energy;voltage converters;chip-favor hardware control;fine-grained power management;adaptive control;model-driven software DVFS manager;sophisticated DVFS-based approach;memory-bound applications;dynamic voltage and frequency scaling;energy efficiency;parallel applications;linear DVFS controller;CoPPer;Copper;Software;Hardware;Timing;Power demand;Computational modeling;Market research;performance;power cap;rapl;control theory;dvfs;self aware systems;adaptive control},
  doi={10.1109/ICAC.2019.00015},
  ISSN={2474-0756},
  month={June},
  ISIArea = {CAS}
}

Downloads: 1

{"_id":"A5mJJ4RnoAQRQqCqu","bibbaseid":"imes-zhang-zhao-hoffmann-coppersoftrealtimeapplicationperformanceusinghardwarepowercapping-2019","author_short":["Imes, C.","Zhang, H.","Zhao, K.","Hoffmann, H."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"propositions":[],"lastnames":["Imes"],"firstnames":["Connor"],"suffixes":[]},{"propositions":[],"lastnames":["Zhang"],"firstnames":["Huazhe"],"suffixes":[]},{"propositions":[],"lastnames":["Zhao"],"firstnames":["Kevin"],"suffixes":[]},{"propositions":[],"lastnames":["Hoffmann"],"firstnames":["Henry"],"suffixes":[]}],"booktitle":"2019 IEEE International Conference on Autonomic Computing (ICAC)","title":"CoPPer: Soft Real-Time Application Performance Using Hardware Power Capping","year":"2019","volume":"","number":"","pages":"31-41","abstract":"Dynamic voltage and frequency scaling (DVFS) has been the cornerstone of innumerable software approaches to meeting application timing requirements with minimal energy. However, recent trends in technology-e.g., moving voltage converters on chip-favor hardware control of DVFS, as hardware can both react faster to external events and perform fine-grained power management across a device. We respond to these trends with CoPPer, which instead uses hardware power capping to meet application performance requirements with high energy efficiency. We find that meeting performance requirements with power capping is more challenging than using DVFS because the relationship between power and performance is non-linear and has diminishing returns at high power values. CoPPer overcomes these difficulties by using adaptive control to approximate non-linearities and a novel gain limit to avoid over-allocating power when it is no longer beneficial. We evaluate CoPPer with 20 parallel applications and compare it to both a classic linear DVFS controller and to a sophisticated control-theoretic, model-driven software DVFS manager. CoPPer provides all the functionality of the sophisticated DVFS-based approach, without requiring a user-specified model or time-consuming, exhaustive application/system pre-characterization. Compared to DVFS, CoPPer's gain limit reduces energy by 6% on average and by 12% for memory-bound applications. For high performance requirements, the energy savings are even greater: 8% on average and 18% for memory-bound applications.","keywords":"energy conservation;multiprocessing systems;power aware computing;soft real-time application performance;hardware power capping;application timing requirements;minimal energy;voltage converters;chip-favor hardware control;fine-grained power management;adaptive control;model-driven software DVFS manager;sophisticated DVFS-based approach;memory-bound applications;dynamic voltage and frequency scaling;energy efficiency;parallel applications;linear DVFS controller;CoPPer;Copper;Software;Hardware;Timing;Power demand;Computational modeling;Market research;performance;power cap;rapl;control theory;dvfs;self aware systems;adaptive control","doi":"10.1109/ICAC.2019.00015","issn":"2474-0756","month":"June","isiarea":"CAS","bibtex":"@INPROCEEDINGS{8831193,\n author={Imes, Connor and Zhang, Huazhe and Zhao, Kevin and Hoffmann, Henry},\n booktitle={2019 IEEE International Conference on Autonomic Computing (ICAC)},\n title={CoPPer: Soft Real-Time Application Performance Using Hardware Power Capping},\n year={2019},\n volume={},\n number={},\n pages={31-41},\n abstract={Dynamic voltage and frequency scaling (DVFS) has been the cornerstone of innumerable software approaches to meeting application timing requirements with minimal energy. However, recent trends in technology-e.g., moving voltage converters on chip-favor hardware control of DVFS, as hardware can both react faster to external events and perform fine-grained power management across a device. We respond to these trends with CoPPer, which instead uses hardware power capping to meet application performance requirements with high energy efficiency. We find that meeting performance requirements with power capping is more challenging than using DVFS because the relationship between power and performance is non-linear and has diminishing returns at high power values. CoPPer overcomes these difficulties by using adaptive control to approximate non-linearities and a novel gain limit to avoid over-allocating power when it is no longer beneficial. We evaluate CoPPer with 20 parallel applications and compare it to both a classic linear DVFS controller and to a sophisticated control-theoretic, model-driven software DVFS manager. CoPPer provides all the functionality of the sophisticated DVFS-based approach, without requiring a user-specified model or time-consuming, exhaustive application/system pre-characterization. Compared to DVFS, CoPPer's gain limit reduces energy by 6% on average and by 12% for memory-bound applications. For high performance requirements, the energy savings are even greater: 8% on average and 18% for memory-bound applications.},\n keywords={energy conservation;multiprocessing systems;power aware computing;soft real-time application performance;hardware power capping;application timing requirements;minimal energy;voltage converters;chip-favor hardware control;fine-grained power management;adaptive control;model-driven software DVFS manager;sophisticated DVFS-based approach;memory-bound applications;dynamic voltage and frequency scaling;energy efficiency;parallel applications;linear DVFS controller;CoPPer;Copper;Software;Hardware;Timing;Power demand;Computational modeling;Market research;performance;power cap;rapl;control theory;dvfs;self aware systems;adaptive control},\n doi={10.1109/ICAC.2019.00015},\n ISSN={2474-0756},\n month={June},\n ISIArea = {CAS}\n}\n\n","author_short":["Imes, C.","Zhang, H.","Zhao, K.","Hoffmann, H."],"bibbaseid":"imes-zhang-zhao-hoffmann-coppersoftrealtimeapplicationperformanceusinghardwarepowercapping-2019","role":"author","urls":{},"keyword":["energy conservation;multiprocessing systems;power aware computing;soft real-time application performance;hardware power capping;application timing requirements;minimal energy;voltage converters;chip-favor hardware control;fine-grained power management;adaptive control;model-driven software DVFS manager;sophisticated DVFS-based approach;memory-bound applications;dynamic voltage and frequency scaling;energy efficiency;parallel applications;linear DVFS controller;CoPPer;Copper;Software;Hardware;Timing;Power demand;Computational modeling;Market research;performance;power cap;rapl;control theory;dvfs;self aware systems;adaptive control"],"metadata":{"authorlinks":{}},"downloads":1},"bibtype":"inproceedings","biburl":"https://bibbase.org/f/nxnupMTqshJBmGaut/cimes-2024-09.bib","dataSources":["FK2Sdqifovg5LjAoQ","8wxz2MZJayyReHAbN","ss7zRWvdAi8FxjJjC","aKv9a6ZaATgxxHGR8","LPA7qi8rmQfuwnQuM"],"keywords":["energy conservation;multiprocessing systems;power aware computing;soft real-time application performance;hardware power capping;application timing requirements;minimal energy;voltage converters;chip-favor hardware control;fine-grained power management;adaptive control;model-driven software dvfs manager;sophisticated dvfs-based approach;memory-bound applications;dynamic voltage and frequency scaling;energy efficiency;parallel applications;linear dvfs controller;copper;copper;software;hardware;timing;power demand;computational modeling;market research;performance;power cap;rapl;control theory;dvfs;self aware systems;adaptive control"],"search_terms":["copper","soft","real","time","application","performance","using","hardware","power","capping","imes","zhang","zhao","hoffmann"],"title":"CoPPer: Soft Real-Time Application Performance Using Hardware Power Capping","year":2019,"downloads":1}