Case Study: a global engineering group forecasts HPC demand and right-sizes core purchases with Focaloid Technologies

A Global Engineering Group predicts HPC demand with Focaloid Technologies, cutting 95th-percentile wait to 2.56 hours

Focaloid Technologies partnered with a global engineering group that was struggling to justify a request for more HPC cores despite an average cluster utilization of only 54%. The challenge was to diagnose why jobs were experiencing long wait times with seemingly low utilization and to provide a data-driven method to accurately price the next hardware purchase.

Focaloid built a discrete-event simulation inside the client's Snowflake warehouse that replayed the actual historical job trace under different growth and core-count scenarios. The solution delivered a Power BI report that accurately predicted a 95th-percentile wait time of 2.56 hours for the following year, which matched the actual outcome exactly. This provided the engineering group with a defensible model to anchor future core purchases to business growth plans, showing that approximately 1,500 additional cores would be required to handle a 25% workload increase while meeting their service target.