An inherent characteristic of IP networks is that they tend to allocate bandwidth as equally as possible among all flows. This was a natural architectural characteristic of the Internet at its onset and served as an equitable mechanism to easily and automatically allocate bandwidth. Unfortunately, aggressive multi-flow applications, peer-to-peer (P2P) being the best known example, have taken advantage of this architectural characteristic by opening up large numbers of flows. A large file transfer, for example, could be divided into ten segments and sent over ten parallel TCP connections. The receiving computer reassembles the segments into a single file and the overall file transfer will complete much more quickly than if the single file had been sent over a single TCP connection. While cleverly exploiting the inherent design of TCP, these applications obtain greater throughput by utilizing more than their fair share of the bandwidth at the expense of other users on the network.
Attempting to correct this imbalance by blocking misbehaving applications—as DPI devices do—is not the right answer. In addition to regulatory concerns about such activity, users themselves are uncomfortable with the idea that service providers are inspecting their traffic and making traffic policing decisions based on its content.
A better approach is simply to task the network with ensuring that all hosts (users or applications) of a have access to the same amount of bandwidth irrespective of how many TCP or UDP flows are initiated.
