The social compatibility (SC) is one of the most important parameters for service robots. It characterises the interaction quality between a robot and a human. In this paper, we first introduce an open-source software-hardware integration scheme for socially-compliant robot navigation and then propose a human-centered benchmarking framework. For the former, we integrate one 3D lidar, one 2D lidar, and four RGB-D cameras for robot exterior perception. The software system is entirely based on the Robot Operating System (ROS) with high modularity and fully deployed to the embedded hardware-based edge while running at a rate that exceeds the release frequency of sensor data. For the latter, we propose a new human-centered performance evaluation metric that can be used to measure SC quickly and efficiently. The values of this metric correlate with the results of the Godspeed questionnaire, which is believed to be a golden standard approach for SC measurements. Together with other commonly used metrics, we benchmark two open-source socially-compliant robot navigation methods, in an end-to-end manner. We clarify all aspects of the benchmarking to ensure the reproducibility of the experiments. We also show that the proposed new metric can provide further justification for the selection of numerical metrics (objective) from a human perspective (subjective).