We conducted in this paper an experimental study on BitTorrent (BT) download performance. Although BT has been widely deployed, many important features about it are not clear yet. For example, what parameter configuration is Optimal for its download performance and to what extent is its incentive mechanism efficient in preventing free-riding. We focused our study on two aspects in this paper: the impact of maximum concurrent upload number on download speed; the efficiency of BT's incentive mechanism on fairness and download performance in the face of free-riders. Our results reveal that there is not an uniform value for the optimal maximum concurrent upload number, and that BT's incentive mechanism is robust to free-riders at a large scale, however seed's upload capacity should be good engineered and BT's incentive mechanism deserve to beimproved in some aspects, e.g., optimistic unchoke.