예측속도 향상 https://cloud.google.com/architecture/minimizing-predictive-serving-latency-in-machine-learning?hl=ko
