Minimizing network traffic is an important issue in content services delivering large volume content files to lower the cost charged for bandwidth and the network infrastructure. Traffic localization is an effective way of reducing network traffic. Peer-assisted content delivery network (CDN) localizes network traffic when a client can obtain the requested content files from a near-by altruistic client instead of the source server. To localize traffic effectively, content files likely to be requested by many clients should be cached locally. This paper presents a novel traffic engineering scheme for peer-assisted CDN models. The key idea is to combine content files while keeping the price equal to the single-content price to induce altruistic clients to request the desired files to be cached. In this paper, we come up with a solution for determining contents combinations that localize download traffic in a network and discuss its upper-bound performance.