1 of 46

Computational Research Facility (CRF) Data Efficiency Analysis

Heather Anderson

Mentor: Charles Liles

Summer 2023 Internship

Langley Research Center (LaRC)

2 of 46

Agenda

Introduction

What is the Challenge?

About LaRC CRF

Project Background

About Dataset

Data Construction

Equipment Terms

About Looker Studio

Data Visualization

Dashboard Benefits

Future Work

Conclusion

3 of 46

Heather Anderson�Introduction

  • Brigham Young University (BYU):
    • Bachelors in Information Systems
    • Expected Graduation: Dec 2024

  • Previous Work Experience:
    • Boeing (2022-2023)
    • StormwaterGO (2022)

4 of 46

What is the Challenge ?

Previous lack of data collection and visualization for the Computational Research Facility

Over-complicated power usage efficiency reports that are difficult to analyze over time

Inability to analyze trends in the data in order to improve facility efficiency

Focus of the project:

Analyze, engineer, and visualize the given data into a more usable format

5 of 46

About the �Langley Computational Research Facility

  • Named after Katherine G. Johnson, NASA mathematician for orbital mechanics
  • $23-million, 37,000-square-foot (3,437 square-meter) structure, consolidating four Langley data centers
  • Opened September 2022
  • The building incorporates energy-saving features that are expected to be 33 percent more efficient than if those features had not been included.

6 of 46

Project Background

Initialized during Spring 2022 to collect data to calculate CRF power usage efficiency (PUE)

3-month process to tag equipment for recording purposes, in order to collect a full year of data to detect changes in PUE

After this data was collected the next step was to find a platform to host the data, PowerBI was not an option because it couldn’t set up automatic data gateways, Tableau was also not an option with the timeline. In the end Looker Studio was selected as it automatically connects to BigQuery and requires low mantinence

7 of 46

Data Pipeline

BAS train ensemble

Aveva PI Data Historian

Python Processing

Big Query

Looker Studio

8 of 46

About the Dataset

  • Data is first processed in Python then sent to BigQuery and served out to Looker Studio in the Cloud environment
  • Big Query is a cloud data warehouse system that allows you to analyze your data at scale
  • Running this data is very inexpensive (Incurred cost of ~4 cents throughout internship)
  • Data Partitioned and engineered by mentor, Charles Liles
  • Pulled directly into Looker Studio and updated every 24 hours

  • Database partitioning (also called data partitioning) refers to breaking the data in an application’s database into separate pieces, or partitions. These partitions can then be stored, accessed, and managed separately.

9 of 46

Previous Reports

10 of 46

Previous Reports

Manual Reports were both error prone and time consuming

11 of 46

Data Construction

Original Dataset

Reconstructed Data

12 of 46

About Power Usage Effectiveness (PUE)

Power usage effectiveness (PUE) is a metric used to determine the energy efficiency of a data center. PUE is determined by dividing the total amount of power entering a data center by the power used to run the IT equipment within it. PUE is expressed as a ratio, with overall efficiency improving as the quotient decreases toward 1.0 (1.10-1.40 is preferable)

Data center infrastructure and the processing power within it require a lot of energy, and data centers that do not operate efficiently will use more energy. Monitoring a metric like PUE is useful for benchmarking data center efficiency. Organizations and data center managers can use this metric once to measure their data center efficiency and then again to measure the effect of any changes made to the data center. This helps reduce power consumption and energy costs

Calculated by Total Facility Power/IT Equipment Energy = PUE

13 of 46

How to Lower PUE

Virtualize servers

Virtual machines can run their own workloads, which reduces energy consumption and frees up more floor space.

Improve cooling systems

To prevent overheating, data centers require a cooling system. However, refrigerant-based cooling systems use a lot of power. Improving these systems or reducing the data center's reliance on them can help lower PUE.

Optimize cool air production

This can be done, for example, by using naturally cool outside air or heat exchangers.

Replace inefficient hardware

The quality and performance of some hardware may degrade over time, so if servers or storage systems are not performing properly, they should be replaced.

Use an energy-efficient uninterruptible power supply (UPS)

Power distribution should be designed with a UPS to be more efficient. More efficient equipment and making power run a shorter distance increase efficiency.

Use energy-efficient lighting

Although lighting generally makes up a smaller portion of power consumption, it is still an easy way to reduce power and heat production. Replacing fluorescent lighting with LEDs on motion sensors and lighting controls can help reduce power consumption and heat production.

14 of 46

Facility Equipment Terms

    • RPPs are able to place a power connection closer to your equipment, they also serve cooling needs and can run while Power Distribution Units (PDUs) may be down for maintenance.

RPP’s – Remote Power Panels

    • An air handler, or air handling unit (often abbreviated to AHU), is a device used to regulate and circulate air as part of a heating, ventilating, and air-conditioning system

AHU – Air Handling Unit

    • In a data center, power meters are typically installed to monitor mechanical loads (ex: pumps, chillers), PDUs (mains and branch circuits), remote power panels (RPPs) (branch circuits), IT busway (feeders and plug-in units), and IT Panel boards (mains & branch circuits).

Power Meters

    • "Critical" power or "IT load" often refers to the data center load that is consumed or is dedicated to IT equipment such as servers, storage equipment and communications switches and routers. Power for lighting or cooling the data center is excluded from "critical" power.

IT Load

    • Data center components require significant infrastructure to support the center's hardware and software. These include power subsystems, uninterruptible power supplies (UPS), ventilation, cooling systems, fire suppression, backup generators, and connections to external networks.

Facility Load

15 of 46

Facility Equipment Terms

  • Cooling Systems:
    • Air cooling is a standard method of system cooling used to the method of dissipating heat. The object being cooled will have a flow of air moving over its surface. Most air-cooling systems use a combination of fans and heat sinks, which exchanges heat with air.
    • Liquid Cooling is the first method within liquid-based cooling which is “water-cooled racks.” This method uses water to cool along the hot side of the cabinet bringing the temperature down. The water is confined within basins and flows from tower pumps through pipes alongside the servers, but does not touch the components of the servers.

16 of 46

About

  • Google’s visualization software, connected to Google Cloud
  • Allows you to create charts with interactive with viewer filters and date range controls

Connect to Data From:

  • Databases, including BigQuery, MySQL, and PostgreSQL
  • Google Marketing Platform products, including Google Ads, Analytics, Display & Video 360, Search Ads 360
  • Google consumer products, such as Sheets, YouTube, and Search Console
  • Flat files via CSV file upload and Google Cloud Storage
  • Social media platforms such as Facebook, Reddit, and Twitter
  • Blended data from any combination of related sources

17 of 46

  • Looker Studio has a wide variety of charts and filters available to visualize your data
  • When you create a chart, it only shows you data that can be visualized in that specific format

Chart Types and Filters

18 of 46

  1. Dashboards can be easily copied with a new data source to maintain report format
  2. Sharing abilities include scheduled report deliveries, embedding capabilities, and google sharing
  3. Version history allows for easy editing and recovery

Additional Development Tools

1

2

3

19 of 46

Data Aggregation

Aggregation is the process of reducing and summarizing tabular data

Aggregation in Looker Studio

There are several ways to apply an aggregation method to your data in Looker Studio:

  • In the data source. A field's default aggregation determines how that metric is displayed in charts. See below for the available default aggregations.

  • In a chart. Report editors can override the default aggregation and apply a different one to the metric on a chart-by-chart basis. Learn how to add and edit data in charts.

  • In a calculated field. You can use specific aggregation functions within a calculated field formula to produce aggregated metrics. See the list of functions.

20 of 46

Visualizing Main 3 Metrics�Page 1

21 of 46

AHU & Real Power�Page 1

22 of 46

Scorecards for Main Dashboard�Page 1

23 of 46

Main Dashboard�Page 1

  • Displayed Metrics:

  • Overall PUE
  • IT Load kW
  • Facility Load kW
  • AHU-1 kW
  • CP-1-1 kW
  • LP-2 kW
  • MDP-2 kW
  • MDP-3 kW
  • RP-2-1 kW

24 of 46

Facility Load vs IT Load�Page 2

25 of 46

Facility Load vs IT Load�Page 2

  • Displayed Metrics:

  • Overall PUE
  • IT Load kW
  • Facility Load kW

26 of 46

Facility Load and Power Meters�Page 3

27 of 46

Facility Load and Power Meters�Page 3

28 of 46

Facility Load & Power Meters�Page 3

  • Displayed Metrics:

  • Overall PUE
  • Facility Load
  • MSB-A
  • MSB-B
  • MSB-C
  • MDBP-1
  • VPM

29 of 46

IT Load and Power Meters�Page 4

30 of 46

Facility Load & Power Meters�Page 4

  • Displayed Metrics:

  • Overall PUE
  • Facility Load
  • MSB-A
  • MSB-B
  • MSB-C
  • MDBP-1
  • VPM

31 of 46

Cooling Tower Fan & �Chiller Plant Cooling�Page 5

32 of 46

Condenser & Chilled Water Pumps�Page 5

33 of 46

Chillers, Pumps, & Fans�Page 5

  • Displayed Metrics:

  • Cooler Tower Fans
  • Chiller Plant Cooling
  • Condenser Pumps
  • Chilled Water Pumps

34 of 46

RPP’s�Page 6

35 of 46

RPP’s & IT Load�Page 6

  • Displayed Metrics:

  • RPP’s
  • IT Load

36 of 46

RPP’s�Page 7

37 of 46

RPP’s & IT Load�Page 7

  • Displayed Metrics:

  • RPP’s
  • IT Load

38 of 46

Power Meters�Page 8

39 of 46

Power Meters�Page 8

  • Displayed Metrics:

  • Overall PUE
  • MSB-A
  • MSB-B
  • MSB-C
  • MDBP-1

40 of 46

Generating Reports

Example: Monthly Report for July 2023 in an Adobe PDF format

41 of 46

Report View

Example: Monthly Report for July 2023 in an Adobe PDF format

42 of 46

Alternate Light Theme

This copy was generated for those who prefer a lighter theme for potentially easier data visibility

43 of 46

Benefits of PUE Dashboards

Fully automated system, replacing previous time intensive manual reports

Low cost to run data (~4 cents throughout internship)

Minimal dashboard maintenance needed; all versions saved to the cloud

Increased ability to analyze data trends over time, and compare metrics over larger periods

44 of 46

Future Work

Utilize this data to improve future power usage efficiency for the data center

Utilize

Potentially integrate this data into the GIS system to make it more accessible

Integrate

Continue to generate reports and edit dashboards according to need

Continue

45 of 46

Conclusion

  • Thank you to the following people and organizations:
  • Mentor: Charles Liles
  • Project Leads: Mark Pugh, Debra Washington, Blain Lege, and Charles Liles
  • LaRC Center Operations Directorate
  • Langley Transformation Initiative
  • NASA Office of STEM Engagement
  • Marshall Space Flight Center (MSFC) Cloud and Computing Services (CCS) Google Cloud Platform (GCP) environment
  • LaRC Integrated Operations Center: Kaitlin Menzies and James Miller

46 of 46

Any Questions?�