In peer-to-peer video-on-demand (P2P VoD) streaming services, a video server load is reduced by peers who cache data of the viewed videos and send these data to other peers instead of the video server. Although FIFO is typically used as the caching algorithm, it is not efficient for using upload bandwidth of the peers because the peers can cache the data of unpopular videos and cannot be requested these data from other peers. In this paper, we propose video-popularity-based caching (VPC) that reduces the video server load by considering video popularity. VPC increases the number of peers who deliver the videos to other peers because the peers give more priority to the data of popular videos than the unpopular videos for caching. In addition, we evaluate the VPC through computer simulations and we show that VPC is efficient for reducing the video server load under steady state condition.