Peer-to-Peer (P2P) systems have brought two serious problems for Internet Services Providers (ISPs): traffic surges and network congestion. Proxy caching for P2P traffic is an effective means of easing the burden imposed by P2P traffic on ISPs. The deployment strategy of proxy caches has a significant impact on their effect on ISPs. It has two important components: one is cache operation that is what P2P traffic caches monitor and serve; the other is cache deployment algorithm that is where to deploy caches. In this paper, we propose a deployment strategy of proxy caches, called LHCDS (Cache Deployment Strategy for Lower and Higher networks), including a new type of cache operation and its corresponding deployment algorithm. It can serve P2P traffic not only from the local lower access network to the higher transit ISP's network but also within the higher transit ISP's network itself. In LHCDS, caches are deployed at or near core routers of the transit ISP. We develop several deployment algorithms based on the new cache operation. We then evaluate the algorithms, and conclude that the proposed greedy algorithm is a viable and practical algorithm whose performance is close to optimal. Experiments also show that ISPs can achieve more P2P traffic decrease on links using LHCDS than previous strategies.