IT and Climate Change: A Startup Builds for a Greener Future
PARIS — What’s a key advantage of Kubernetes? It accommodates scaling. In other words, it makes it easier to use more compute, more electricity, more resources.
“This entire ecosystem of cloud native Kubernetes, it’s all geared up to increase consumption. While, especially in Europe, there is a demand to decrease,” Pini Reznik, CEO and co-founder of re:cinq, told The New Stack while taking a break from his company’s booth at KubeCon + CloudNativeCon Europe.
In the United States, politicians still argue about whether climate change is real. Here in Europe, however, lawmakers are actively fighting it. The European Commission’s Fit for 55 program sets out legally binding targets for decreases in carbon emissions: a reduction of 55% by 2030. Its overall Green Deal initiative aims to make the European Union’s member countries carbon neutral by 2050.
And that, said Reznik, presents both an incentive for organizations to seek more sustainable IT practices — and business opportunities. His startup is creating a platform designed to reduce carbon emissions from IT operations.
“It feels like five years ago, people cared — but now, they care and are willing to spend on it,” he said. “There is a feeling that things are actually happening. That’s why it’s important for us that it’s not an altruistic thing, but that it is now a viable business. I’m going to believe that it will become a massive industry.”
Measuring Emissions in Real Time
Re:cinq’s co-founders — Reznik, Kristian Ladefoged and Michael Mueller — shared a background in consulting, working together at Reznik’s previous company, Container Solutions, which advised enterprises on cloud native transformation projects.
Their new company, founded in October, didn’t start out intending to build a product. As it began seeking its first clients, it discovered that organizations needed more than just good advice and the latest GreenOps techniques.
“We started thinking, we will go as consultants to their company and will improve their system, so they will be emitting less emissions,” Reznik said. “OK, but how do we know how they make now, so we can later say that we saved that much?”
What re:cinq’s consultants discovered, he said, is that “there is no practical way to measure. So how can you improve if you can’t measure?”
Often, Reznik said, “savings” are measured in currency, not carbon. While public cloud vendors offer customers some information on their carbon emissions, it’s infrequent, it’s monthly, it comes after the fact. It’s not real-time,” he said. “So it’s OK for financial reporting to the government. But it’s not relevant if you want to take action and see how the system is actually going up and down.”
Given that there was no practical way to measure the impact of an IT team’s actions on its emissions in real-time, re:cinq started building a tool. Right now it’s called “Cloud Carbon,” though Reznik said that name is a work in progress.
Cloud Carbon, an OpenTelemetry exporter, collects basic information from a cloud or server, showing the emissions generated by an organization’s cloud infrastructure. It is under an Apache 2.0 license.
Re:cinq, Reznik said, will most likely donate Cloud Carbon to the Cloud Native Computing Foundation. “This is something that everyone should have access to,” he said. However, he added, the platform his company is building will remain proprietary.
Cloud Carbon is open source, and looking for contributors. “We already heard quite a few people here, willing to contribute,” he said.
“The idea is really to give as comprehensive a state of carbon emissions as possible, not just cloud or on-prem, and potentially also a very granular approach. Like. you want to know which application, even if it’s running in different places, how much emissions it actually produces.”
AI and Carbon Consumption
Of course, even as climate change is having its way with the planet, the tech industry is now in a fever for all things AI. Training large language models (LLMs) takes a staggering amount of computing resources and electricity.
Reznik questions whether that’s a given, however. “Do we really need to train those models from scratch? Or maybe we can share open source models? And maybe we can train on smaller datasets? Maybe we can use greener energy, be carbon aware and things like that.”
The question he asks of the AI revolution is “How can we get the value without destroying the planet?”
More discussion of sustainability in software production is essential, he said. “One of the biggest problems I see right now is that there is no awareness. Yes, we want to be green. But what does it mean?“
The keynote presentations on KubeCon’s first day were entirely focused on AI; the second day’s keynotes, Reznik noted, included a couple of sustainability-focused talks. “I feel it could be a lot more,” he said.
The cloud native ecosystem celebrated by KubeCon, he said “is based on almost unlimited growth.” But that growth, he added, needs to be more thoughtful.
“I’m not saying, don’t sell your stuff to your customers,” Reznik said. “It’s more like, do you really need 100 servers, or you can manage exactly the same thing with five?”
Demand for data centers is expected to grow 10% a year until 2030, according to a report released by McKinsey and Company in January 2023. However, the analyst also reports only about a 15% average rate of data server utilization; other studies peg that figure anywhere between 12 to 18%.
That much waste “doesn’t sound smart,” Reznik observed.
He added, “Even when the servers are idle, there is embodied carbon, they still consume power. And it costs money. So why are we doing this?”
