Company
Mercedes-Benz.io develops software and technology solutions for Mercedes-Benz’s digital platforms. Their main products are the Mercedes-Benz website, e-commerce platform, digital services, and aftersales solutions.
Challenge
The team faced several scaling and efficiency challenges in the Kubernetes environment. Workloads that were not rightsized led to unnecessary resource consumption, and the use of outdated instance types caused the team to miss out on the performance and cost benefits offered by newer generations. The large scale of Mercedes-Benz.io’s Kubernetes footprint only amplified these inefficiencies, resulting in significant operational overhead and cost.
Solution
To address these challenges, Mercedes-Benz.io turned to Cast AI’s automation platform, implementing automated instance selection, node autoscaling, workload rightsizing, and other resource optimization strategies to align capacity with actual demand.
By using modern instance types and tailoring node sizes to fit specific product requirements, the team significantly improved performance while reducing costs and manual engineering effort.
Adopting intelligent workload scheduling and continuous monitoring further ensures efficient resource utilization, enabling the company’s Kubernetes environment to scale more efficiently while minimizing waste and operational overhead.
Results
- Significant cost savings through smart bin-packing efficiency and Spot VMs
- More efficient node scaling and intelligent instance selection
- Eliminating complexity and lowering operational overhead for Kubernetes cluster management
Rebalancing and cluster bin-packing
Rebalancing allows a cluster to reach its most optimal and up-to-date state—suboptimal nodes are automatically replaced with new ones that are more cost-efficient and run the most up-to-date node configuration settings.
The Mercedes-Benz.io team runs a regular rebalancing operation, where new VMs are provisioned and pods are bin-packed to maximize resource utilization and reduce costs.
Workload rightsizing
Workload Autoscaler is a Cast feature that automatically scales workload requests up or down to ensure optimal performance and cost-effectiveness.
Spot VM automation
The Cast autoscaler lets users like Mercedes-Benz.io run workloads on Spot VMs. By automating the entire Spot lifecycle – from provisioning to scaling and decommissioning – Mercedes-Benz is able to generate significant cost savings.
The graph below shows that manual management of Spot VMs and on-demand instances (from December 5 to 19) is suboptimal when compared to the previous period, where workloads ran 100% on Spot VMs.
Node hibernation
Cast AI shrinks worker nodes to zero when resources are not needed to boost efficiency and reduce costs.
Using Spot VMs wasn’t in our initial scope; we focused on improving the general scaling behavior of our cluster. But the Cast AI platform offers easy implementation of using Spot VMs – in the end, they’re just another feature toggle in the node template definition.
From the platform team perspective, this allows us to give your product teams the possibility of using Spot Instances while being in full control of the workload, for example, by setting tolerations and node selector configurations for Spot.
This was a quick win for us, especially when it comes to the cost. It was really easy for us to enable teams to use Spot and save money along the way.
Bertram Hass, CloudOps Engineer at Mercedes-Benz.io
