Forging a path to real estate sustainability

The real estate sector represents 35% of energy-related carbon emissions in the European Union. Ampeers Energy aims to reduce this number through a combination of cutting-edge intelligent software and hardware. Headquartered in Munich, Germany, the company develops optimal energy concepts and automates processes to help real estate customers reduce CO2 emissions by up to 90% and increase their revenue.

To support Ampeers Energy in its mission, the infrastructure team needs to meet several key requirements. “Our infrastructure needs to be highly available to gather as many data points as possible from our distributed systems, with many ingress points that need to be processed individually. We use the data to run simulations, create predictions about the energy usage of a district, and carry out optimizations based on all that,” said Lukas Marquardt, DevOps Engineer at Ampeers Energy.

Lack of cost visibility has become a mounting concern

As the company’s cloud footprint across Azure services grew, the infrastructure team began looking for a solution that would provide a view into cloud expenses at a level of granularity that made sense both for the company’s use cases and its core cloud-native technology, Kubernetes. 

The team initially experimented with the open-source cost monitoring tool OpenCost. However, the setup turned out not to be user-friendly and the metrics delivered weren’t accurate for the particular use cases at Ampeers Energy. The team kept looking for alternative solutions that would be easier to onboard and provide more accurate metrics. 

At the beginning, we were looking for a solution that could show us which producer generated which cloud costs –  and how to optimize that. I came across Cast AI, where I could onboard quickly and connect my cluster to instantly see a cost overview.

Lukas Marquardt, DevOps Engineer at Ampeers Energy

Lukas set it all up and gave me the login to the Cast AI platform so I could see the cost visibility features for myself. My impression was that all the data was very accessible. Since there’s no need for lengthy onboarding or understanding the product, this was really a no-brainer.

Lucas Recknagel, CTO at Ampeers Energy

The cost optimization journey started with monitoring

Ampeers Energy implemented Cast AI and instantly got insights into the actual resource utilization across their Kubernetes cluster.

The first thing we noticed was the clear separation of the different namespaces and workloads in our cluster. Having that in front of us, we could easily see where our main problems are – and start solving them to optimize our setup. We could see where there were no resource limits or requests set and immediately fix that. It was just a quick rundown on where we could make some quick wins.

Lukas Marquardt, DevOps Engineer at Ampeers Energy

The highly accurate cost and utilization tracking also helped the team reduce node size and number, driving further savings. 

Once we defined the resource limits, we saw that we could reduce our node count. Since implementing Cast AI, we have learned that our nodes have a great variation in utilization. We had large nodes with 90% utilization and smaller nodes half their size but utilized by only 10%. With Cast AI, we could achieve a smarter pod distribution that led to significant cost savings.

Lukas Marquardt, DevOps Engineer at Ampeers Energy

33% cost savings with one click of a button

Before turning on the automated cost optimization in Cast AI, Ampeers Energy had to solve a few configuration matters in their infrastructure, such as adding Pod Disruption Budgets. A Cast AI engineer assisted the team throughout this process. “The production started with a Solutions Architect and he was great to work with. He is a very experienced engineer who greatly helped us in optimizing our cluster to industry-based best practices and preparing everything for automated cost optimization,” said Lukas Marquardt.

After we configured everything to be in line with industry best practices, it was really just a single meeting. We pressed a button, waited for 30 minutes, and after that, everything was done. We saw a 33% cost reduction on the cluster connected to Cast AI.

Lukas Marquardt, DevOps Engineer at Ampeers Energy

Thanks to the team’s involvement with the cost monitoring module, the ongoing cost optimization by Cast AI is completely transparent. “We have a glass pane for everything. We can see the cost, the container security best practices, and a handy dashboard. I controlled our expenses as we turned the automated cost optimization on and could see how our workload scaled and how our costs went down,” said Lukas Marquardt.

The autoscaling features of Cast AI had a positive impact on the engineers’ workload as well.

We not only adjusted our cluster and workloads to industry standards but also gained the peace of mind of no longer having to do anything manually about cluster scaling. If we reduce our workloads, we can trust Cast AI to reduce our node count. And if we increase our workloads, the opposite will occur. Cast AI makes everything around Kubernetes easier.

Lukas Marquardt, DevOps Engineer at Ampeers Energy

Next steps: driving further cost reductions with spot instances

Ampeers Energy is planning to connect more clusters to Cast AI. “This is our first cluster, the development environment. We’re planning to onboard our staging and production environments as well. So, we expect to achieve more savings in the near future,” said Lukas Marquardt.

The team also aims to leverage spot instance automation to achieve even higher cost savings.

We aren’t using spot instances right now because our workloads aren’t optimized for that. But we tested a duplicate of our cluster to see what the cost savings would amount to if we turned spot instance automation on. We found out that we could save 40-50% on top of the 33% we’re currently saving. We’re very optimistic about our future collaboration with Cast AI.

Lukas Marquardt, DevOps Engineer at Ampeers Energy