Not All Federated Learning Algorithms Are Created Equal: A Performance Evaluation Study