High Performance Computing (HPC) facilities play a crucial role in acceler- ating Electronic Design Automation (EDA) procedures at NXP Semiconduct- ors. The increasing number of job requests and workloads has led to a surge in memory demand, which is a costly resource. To address this, we leverage memory swap space on disk as a more affordable extension of main memory. By employing a job submit-time parameter called memory reservation, we can increase the cluster jobs’ parallelism, and control the memory swap space usage. The primary goal of this thesis is to find the optimal cost and performance by optimizing the memory reservation parameter. To achieve this goal, we created an effective parameter optimizer that employs a one-way search method to determine the best memory reservation, focusing on scenarios with a large homogeneous job set running on an overloaded system. Our contributions include improving the memory utilization efficiency in HPC by utilizing swap space. Previous work used Machine Learning models to predict maximum memory usage and reduce over-reservation; and our work further reduces memory usage per job. The results demonstrate that the parameter optimizer effectively minimizes the cost and performance at the same time, leading to up to 43% cost saving for a job among our test cases.