How to Reduce the Carbon Footprint of Your Data Center?

Listen to the podcast

Abstract in English

Let's start with getting to know you better, Bertrand. Tell us about you and your professional background.

Well, I'm a self-taught developer. I started coding at 13 on the AMSTRAD computer my mother offered me. I instantly became a software programming aficionado! I studied and graduated in system engineering and rapidly started to work as a consultant on BMC PATROL. As I built a more profound knowledge of the product, I became interested in designing my monitoring software to close the gaps I had identified in PATROL. To make this project real, I've created Sentry Software.

Tell us about Sentry Software, how it started, and your current positioning?

To make it simple, we started with a monitoring module that would collect hardware information from different types of servers regardless of their brand… “A module to collect them all”!!! We specialize in hardware monitoring to help companies detect and prevent hardware failures and poor performance, ultimately avoiding downtime and availability issues. We had the chance to close a partnership deal with BMC: our product would be sold with their platform, and it started being implemented in major data centers worldwide right from the beginning. We are focused on R&D operations; we do not have sales representatives; this is BMC's job. We are a relatively small French company of 30 collaborators, but our products target large organizations, such as telecom companies, governmental organizations, banks, etc.

What is the current situation regarding exposing hardware health and performance metrics?

Over the years, many standards have been used to collect hardware I/Os, health, and performance information. Each manufacturer tries to implement a standard, usually in a partial way, and tries to compensate for imperfections with other technologies. Today, everything is supposed to be managed with REST APIs; it is the next best thing, the Eldorado of observability. But some basic REST principles are often ignored, resulting, again, in imperfect solutions. This is very challenging for organizations like ours. This being said, OpenTelemetry brings great hope as it could become the universal instrumentation standard for capturing metrics.

Our goal with Hardware Sentry is to collect any hardware health and performance data from any server and expose normalized metrics regardless of the manufacturer or the device model. We applied this principle to sustainable key indicators, such as electricity consumption and carbon emissions.

We added environmental considerations to our observability solution starting in 2008 (Green IT). Organizations concerned by their carbon footprint, primarily European companies, need help finding efficient solutions. Evaluating the carbon footprint of a data center is challenging, as it includes servers, terminals, but also operating and managing aspects, and now cloud and SaaS services. How can we determine the carbon footprint of all these elements? Well, we measure the electricity consumption of servers and other hardware devices, multiply these values by the related carbon emission known for these components, expose a carbon footprint by device/component, and offer areas for improvement.

Why OpenTelemetry, and what is your position about open source?

Historically, we are a company that creates proprietary products (old school). We exclusively designed software that integrated with BMC platforms. But times are changing, and we wanted to open up to more recent and even trendy ecosystems (Prometheus, Zabbix, etc.). We discovered OpenTelemetry, “the” emerging standard in observability that defines its communication protocols, data format, and metrics, logs and traces naming conventions. While recent, OpenTelemetry has already been adopted by major observability players, Splunk, Datadog, Prometheus, Grafana, New Relic, Dynatrace, but also AWS, Google, and Microsoft Azure, who are engaged in the definition of this standard. It is here to stay! And yes, it is fully open source! So, we jumped in and became contributors to OpenTelemetry. This made us seriously consider going open source with our flagship product Hardware Sentry. We are in a transition phase, managing legal and technical challenges. We expect to be ready to go fully open source by the end of 2023.

To wrap this up, what is your vision of the future challenges in the observability ecosystem?

Future challenges will consist in reducing IT power consumption. Until now, we relied on Moore's Law, assuming we can do more without blowing budgets. So, we added security, backups, anti-virus, and… observability. But tomorrow, we will face choices because financial investments have limits. For example, we may have to question the necessity of keeping years of bank data. We will need to put the environment first!

About Tech.Rocks

Tech.Rocks is the 1st community of Tech professionals in France. It promotes Tech Leaders messages through communities, contents, and events. Visit https://www.tech.rocks/ for additional information.