A vehicle in permanent Sport mode consumes more fuel than necessary. The same applies to cloud platforms: the operating model and computing power must match the actual workload. ITM Consulting therefore realigned the Databricks environment for the SAP LeanIX integration of a leading German automaker. Within five months, monthly costs dropped from around 24,000 to 123 U.S. dollars.
01 Background
Functionally reliable, but not economically optimized
Databricks handled integrations, automations, scripts, and webhooks for an SAP LeanIX environment with 22 fact sheet types and more than 240,000 fact sheets. From a functional standpoint, the environment worked as intended. However, there was no overarching strategy for determining which workloads actually needed to be run on Databricks, how they could be bundled, and how much computing power they actually required. This resulted in monthly costs of approximately $24,000.
The customer had set an annual savings target of $150,000. To achieve this, simply configuring individual clusters to be smaller was not enough. First, it was necessary to determine which processing tasks SAP LeanIX could handle on its own. Once that was established, the remaining Databricks capacity could be sized appropriately.
02 Procedure
First, migrate workloads; then, optimize compute resources
Phase 1: Migrate workloads back to SAP LeanIX
ITM Consulting reviewed all integrations, automations, scripts, and webhooks. Tasks that could be handled using native SAP LeanIX features—such as automations, calculations, or scheduled connectors—were migrated back from Databricks to SAP LeanIX.
The remaining workloads were consolidated. Where it made business sense, scheduled executions replaced event-driven processing. Streamlined trigger events and payloads also reduced the number and scope of processing operations. This first phase reduced the original costs by 77%.
Phase 2: Reduce remaining Databricks costs by an additional 98%
For the remaining Databricks workloads, ITM Consulting adjusted the DBU usage and cluster configuration to match the actual workload. Computing power was reduced to the lowest level that ensures stable operation.
Right-sizing reduced the costs remaining after Phase 1 by an additional 98%. The total reduction from both phases amounts to 99.5%.
Continuously monitor cost trends
Databricks dashboards provide ongoing visibility into usage and costs. Discrepancies become apparent early on, and the environment can be readjusted to meet new requirements. This ensures that cost savings are sustained over the long term.
The greatest savings did not result from smaller clusters alone. The key was to avoid unnecessary Databricks workloads and to precisely size the remaining computing power.
03 Result
From about 24,000 to 123 U.S. dollars per month
After five months, the monthly Databricks costs totaled $123. Compared to the initial figure of approximately $24,000, this represents a reduction of 99.5%. If this cost level remains stable, it will result in potential annual savings of approximately $287,000.
The architecture today is also streamlined and straightforward. SAP LeanIX handles all tasks that can be natively processed within it. Databricks is reserved for workloads that require external computing power. This reduces complexity and simplifies ongoing cost management.
Would you like to structurally reduce the costs of your cloud environment? Talk to us about your opportunities for optimization.