Blog

Sentry Software Reports its Data Center Efficiency Live

A step forward on IT sustainability

Sentry Software Reports its Data Center Efficiency Live

Reporting real time data on the efficiency of data centers is no longer reserved to the top cloud providers (Facebook, Google, AWS, Azure). More modest companies can do the same with only a few tools and resources. In this article, we will tell you how we did it at Sentry Software.

Step 1: Choosing the PUE as a reference

Sentry Software uses the Power Usage Effectiveness (PUE), a standard promoted by The Green Grid and others in the data center industry, to assess the energy efficiency of its data center. The following information is required to calculate it: the total electricity power consumed by the data center and the power used only by the IT equipment in that facility:

The ideal PUE is 1, which means that 100% of the data center electricity consumption is going to IT (computation). As illustrated in the graph below, transformers, UPS, lighting, and cooling have a direct impact on the PUE:

Power Usage Effectiveness (PUE) by server space type. Source: Lawrence Berkerley National Laboratory

Power Usage Effectiveness (PUE) by server space type. Source: Lawrence Berkerley National Laboratory

Step 2: Measuring the electricity consumed

The system administrators installed 3 physical sensors (DIRIS from Socomec) at the Paris facility to measure the power consumed by the IT infrastructure and more especially:

  • The cooling system power (PCooling)
  • The UPS input power (PUPS)
  • The servers, network and storage power (PIT)

The collected information is exposed in the Socomec Web interface:

Power consumption reported by the first physical sensor

Power consumption reported by the first physical sensor

Power consumption reported by the second physical sensor

Power consumption reported by the second physical sensor

Power consumption reported by the third physical sensor

Power consumption reported by the third physical sensor

Step 3: Calculating the PUE

Once the sys admins made sure that the power consumption was properly reported, they configured Monitoring Studio X to collect the information from the 3 DIRIS sensors and calculate the PUE of the room based on the values collected for these sensors:

Monitoring Studio collects the information from the physical sensors and calculates the PUE

Monitoring Studio collects the information from the physical sensors and calculates the PUE

The UPS input power sensor includes the power consumed by all the equipment protected by the UPS as well as a small fraction “wasted” by the UPS. We considered an efficiency ratio of 96% for our Socomec UPS, which gives us this formula:

The sysadmins then configured Monitoring Studio X to expose these metrics as a Prometheus exporter:

Monitoring Studio configured as a Prometheus Exporter

Monitoring Studio configured as a Prometheus Exporter

Step 4: Reporting the PUE live

To report the PUE live, the sysadmins configured the OpenTelemetry Collector bundled with Hardware Sentry to scrape the metrics exported by Monitoring Studio X, including the PUE, and push the hw_site_pue_ratio metric to the Prometheus server of our demo platform.

Note: Our demo platform is simply made of a Prometheus server, combined with Grafana Dashboards. Grafana was configured to allow anonymous access and loaded with the Hardware Sentry dashboards.

Note that the sysadmins could have configured our Prometheus server to directly scrape the metrics from Monitoring Studio X, but for security reasons we used the OpenTelemetry Collector to scrape these metrics on the internal network and push the metrics to the outside using the Prometheus Remote Write protocol (our demo platform is hosted on a public cloud).

Architecture diagram of the PUE Live project

Architecture diagram of the PUE Live project

Once the PUE metric was properly displayed in the Hardware Sentry Dashboards for Grafana, we shared and embedded the corresponding panel on the home page of Sentry Software's web site as a simple iFrame.

Sentry Software reports its PUE live on its web site

Sentry Software reports its PUE live on its web site

Step 5: Going further and reporting the electricity costs and carbon emissions live

The Hardware Sentry Dashboards for Grafana rely on the hw_site_electricity_price_dollars and hw_site_carbon_density_grams metrics which are usually set to static values in the Hardware Sentry's configuration file. The calculation of the electricity cost and carbon footprint based only on hardcoded information, proved to be less relevant and accurate in a time of higher and more volatile energy prices.

To obtain that information in real-time, Sentry Software needed to regularly retrieve the price and carbon density of the electricity produced in France. The sysadmins therefore configured Monitoring Studio X to query the French Electricity Transmission Operator (RTE) data and get the carbon density every 15 minutes and the electricity price every hour.

As for the PUE, the sysadmins configured the OpenTelemetry Collector to scrape the metrics from Monitoring Studio X, push them to the Prometheus Server, and have them exposed in the Hardware Sentry Dashboards of our demo platform. This implementation allows us to know in real-time the energy cost and carbon emissions of our data center:

Sentry Software knows in real time the electricity cost and carbon emissions of its data center

Sentry Software knows in real time the electricity cost and carbon emissions of its data center

Architecture diagram of the project

Architecture diagram of the project

Once this information was available in the Hardware Sentry Dashboards for Grafana, we shared and embedded the corresponding panel on the home page of Sentry Software's web site as a simple iFrame.

Sentry Software reports its yearly CO2 Emissions

Sentry Software reports its yearly CO2 Emissions

Step 6: Running the project postmortem

Sentry Software is proud to have been able to report its PUE, electricity costs, and carbon emissions with a few tools:

  • 3 physical sensors to measure the power consumed by its IT infrastructure,
  • Monitoring Studio X to:
    • Collect the information reported by the physical sensors and compute the PUE
    • Retrieve the price and carbon density of the electricity produced in France, where its data center is based.
  • OpenTelemetry to scrape these metrics and push them to our Prometheus server
  • Hardware Sentry to monitor our servers, network switches and storage systems and their energy usage
  • Hardware Sentry Dashboards to expose the information in comprehensive Grafana dashboards.

Reporting real time data matters!

By reporting the PUE and yearly carbon footprint live, Sentry Software can quickly identify the initiatives that have a direct impact on the energy efficiency of its data center and adjust its actions accordingly. Among the many options considered to improve its PUE, Sentry Software plans to gradually increase the temperature of its data center as recommended by Hardware Sentry ™.

Feeling skeptical? Access our demo platform and check out our PUE, electricity costs and carbon emissions live!