Figure 1: Percentage of access traffic from top networks Inefficient routing protocols. Although it has managed admirably for scaling a best-effort Internet, BGP has a number of well-documented limitations. It was never designed for performance BGP bases its route calculations primarily on AS hop count, knowing nothing about the topologies, latencies, or real-time congestion of the underlying networks. In practice, it is used primarily to enforce networks business agreements with each other rather than to provide good end-to-end performance. For example, [34] notes that several paths between locations within Asia are actually routed through peering points in the US, greatly increasing latency. In addition, when routes stop working or connectivity degrades, BGP can be slow to converge on new routes. Finally, it is well-known that BGP is vulnerable to human error as well as foul play misconfigured or hijacked routes can quickly propagate throughout the Internet, causing route flapping, bloated paths, and even broad connectivity outages [25].
