Tensorflow serving performance optimization