It’s become the favorite whipping boy of networking. The Internet is erratic. The Internet is unstable. The Internet is unsecure. But exactly what is wrong with the Internet and can it be fixed?
We dove into that question with our co-founder and CTO Gur Shatz in a recent eBook, “The Internet is Broken: Why Public Internet Routing Sucks.” You can read it for yourself here.
Since the early days of the Internet, routers were shaped by a myriad of technical constraints. General purpose processors lacked the processing power, forcing router vendors to rely on custom hardware. To deliver line rate performance, packet processing was kept to a minimum and routing decision were moved to a separate process, the “control plane.”
With the separation of the control and data planes, architects could build massively scalable routers. Compute resources became more than sufficient for high-speed packet processing.
And yet “feedback” from the data plane to the control plane continues to be nominal, an anachronism from the early days of the Internet.
A modern router has little insight into the packet flows — how long it takes to reach the next hop, the degree of congestion in the network, or the nature of the traffic being routed. To the extent that such information is available, it will not be factored into the routing decisions made by BGP, the routing protocol gluing the Internet together.
MPLS services are not the answer. They’re too expensive and changes take too long for today’s business. Cloud and Internet performance suffer because of traffic backhaul, a common phenomenon in the way company’s architect their MPLS-based backbones. Local loop availability is often an issue with the way MPLS implementation. The protocol also suffers from many of the same problems as BGP.
There are measure that can be taken to address the problem, though. This eBook details those measures and explains how they address the limitations of the Internet and MPLS. To read, the eBook click here.