| NSF Org: |
CCF Division of Computing and Communication Foundations |
| Recipient: |
|
| Initial Amendment Date: | July 27, 2015 |
| Latest Amendment Date: | July 27, 2015 |
| Award Number: | 1527249 |
| Award Instrument: | Standard Grant |
| Program Manager: |
Yuanyuan Yang
CCF Division of Computing and Communication Foundations CSE Directorate for Computer and Information Science and Engineering |
| Start Date: | August 1, 2015 |
| End Date: | July 31, 2020 (Estimated) |
| Total Intended Award Amount: | $369,092.00 |
| Total Awarded Amount to Date: | $369,092.00 |
| Funds Obligated to Date: |
|
| History of Investigator: |
|
| Recipient Sponsored Research Office: |
4000 CENTRAL FLORIDA BLVD ORLANDO FL US 32816-8005 (407)823-0387 |
| Sponsor Congressional District: |
|
| Primary Place of Performance: |
FL US 32826-3252 |
| Primary Place of
Performance Congressional District: |
|
| Unique Entity Identifier (UEI): |
|
| Parent UEI: |
|
| NSF Program(s): | Software & Hardware Foundation |
| Primary Program Source: |
|
| Program Reference Code(s): |
|
| Program Element Code(s): |
|
| Award Agency Code: | 4900 |
| Fund Agency Code: | 4900 |
| Assistance Listing Number(s): | 47.070 |
ABSTRACT
Recent years have seen many well-recognized energy conservation schemes developed for big data computing infrastructures, which aggregate heavy workloads on either a few chips or devices. While both of these methods reduce energy consumption, they can also elevate temperature levels on long standing IT instruments and ultimately cause them to overheat. As a consequence, the reliability of these devices can be significantly degraded as shown in many recent studies. In the worst cases, they can fail or malfunction. There is an imperative need for developing new power and energy control solutions in the multibillion-dollar industry of big data computing.
This research project is directed towards system-level solutions for temperature constrained data center energy management. It will develop methods and tools for controlling energy consumption in high-performance data centers operating under various temperature constraints by exploring the use of fuzzy decision-making techniques. Most of the existing studies rely on a well-developed relationship model between each criterion including open-loop search and optimization methods and rigid control schemes with fixed temperature constraint assumptions. In contrast, this project pursues solutions from a different angle: given that each criterion is in a non-linear relationship with another in a data center, how can new modeless control techniques that do not rely on fixed constraints work successfully? The project, if successful, will achieve foreseeable societal gains in terms of environmental benefits of energy conservation, potential for commercial impact of reducing costs and increasing operational efficiency in both warehouse-scale computer systems and data centers and reducing the data-center temperature to maintain reliability.
PUBLICATIONS PRODUCED AS A RESULT OF THIS RESEARCH
Note:
When clicking on a Digital Object Identifier (DOI) number, you will be taken to an external
site maintained by the publisher. Some full text articles may not yet be available without a
charge during the embargo (administrative interval).
Some links on this page may take you to non-federal websites. Their policies may differ from
this site.
PROJECT OUTCOMES REPORT
Disclaimer
This Project Outcomes Report for the General Public is displayed verbatim as submitted by the Principal Investigator (PI) for this award. Any opinions, findings, and conclusions or recommendations expressed in this Report are those of the PI and do not necessarily reflect the views of the National Science Foundation; NSF has not approved or endorsed its content.
Modern data center becomes more heterogeneous and imposes new computing metrics to deal with emerging big data and big learning applications. Performance, power and energy efficiency, along with temperature and thermal energy are demanding measures to operate a high-performance, efficient data center today. Due to staggering electricity bills, an increasingly large number of energy conserving requirements have been introduced on modern data centers. To meet these requirements, data centers have been considering increasing the set-point temperature at which to run the cooling system. On the other hand, various energy conservation schemes have been developed for multi-core CPUs, storage systems, and servers, to aggregate heavy workloads on a few CPU cores or disks. While these methods reduce energy consumption, they could elevate temperature levels on long standing IT instruments and ultimately cause them to overheat. Therefore, the reliability of these devices could be significantly degraded as shown in many recent studies.
This research project is directed towards system-level solutions for temperature constrained data center energy management. We have developed methods and tools for controlling energy consumption in high-performance data centers operating under various temperature constraints by exploring the use of fuzzy decision making techniques. The approach is based on a simple premise: the trend in modern data center is towards operating under service-level agreement on multiple constraints such as performance, energy-efficiency, reliability, etc. We are investigating how to exploit fuzzy decision making theory to solve this ever-important multi-constraint optimization problem in data centers, in particular to constrict temperature.
The major research outcome is a unified framework in which the entire data center architecture and individual components (e.g., CPU, storage) use a single underlying mechanism to address both reliability and energy conservation concerns. The synergy among dynamic multi-constraint control in data center is well recognized in a number of contexts. Most of the studies rely on a well developed relationship mode between each criteria including open-loop search and optimization methods and rigid control schemes with fixed temperature constraint assumption. Our angle here is different: given that each criteria is in a non-linear relationship with another in a data center, how can we develop new control techniques that are modeless and do not rely on fixed constraints.
More specifically, we achieve the following research outcomes. First, we develop a Fast Multi-Constraint Fuzzy Markov Decision Making Controller for Large Scale Data Center by Using Possibility Theory. We extend the temperature constrained energy efficient controller to large scale HPC environments.
Second, we develop a Dynamic Performance, Energy and Reliability Control for SSD-HDD Hybrid Systems. We develop the controller for SSD-HDD hybrid systems to balance performance, power consumption and reliability.
Third, we develop a Fuzzification Control for CPU DVFS with consideration of I/O Wait and Temperature Constraint. We apply a Fuzzy Decision Making algorithm on Multi-Core CPU task schedulers.
Last Modified: 10/10/2020
Modified by: Jun Wang
Please report errors in award information by writing to: awardsearch@nsf.gov.
