We evaluate the effectiveness of five machine learning (ML) algorithms (Support Vector Machines, XGBoost, Logistic Regression, Random Forest, and Deep Neural Networks) for real-time detection of distributed denial of service (DDoS) attacks, comparing both detection accuracy and computational overhead. Using a dataset of network flows collected from different major cloud providers, the ML algorithms are assessed based on performance metrics like accuracy, recall, precision, and F1 score, while also measuring CPU utilization, memory footprint, and prediction time. The Random Forest model demonstrates the best overall performance with an accuracy of 99.99% and an F1 score of 0.9999, with prediction time of 2.04 microseconds per sample. While XGBoost showed comparable accuracy (99.96%), Random Forest exhibited better precision. In terms of resource efficiency, Logistic Regression achieved notable results with 94.28% accuracy while requiring only 0.038 microseconds per inference and minimal memory usage (36.74 MB during training). Our comparative analysis provides empirical evidence for selecting optimal machine learning approaches for DDoS detection, considering both performance and resource constraints.