Nearly every new network innovation uses what we have dubbed the “Lowest Common Denominator” network (i.e., “middlebox mindset”), featuring 1990’s era networking, to talk to each other. If any new “features” interwork at all, it is only through coordinated configuration or orchestration.
Most of these technologies also utilize non-standard, stateful capabilities (i.e., tunneling, addressing changes, proprietary policies distribution, scalable databases) that require additional pre-shared configuration information. All of this configuration and orchestration greatly expands the surface area for attacks and complicates network understanding and operation.
We really can’t look at the large internet players to make things easier on the networking front. Most recent technology giants have developed networking technologies that cannot be shared or interconnected with each other. Over time, this type of thinking has perpetuated major weaknesses and scaling issues that need to be resolved in networks today. Let’s get some perspective:
- Facebook developed its own LISP-LIKE scheme by splitting 128-bit IPv6 addresses into two parts — location and service. Facebook also developed switch control software that is open sourced for routing and switching inside data centers and is now developing Route R to augment internetworking between data centers.
- Amazon developed Amazon Web Services utilizing a proprietary database scheme that stores Layer 2 MAC addresses.
- Google developed its own routers and technology for improving routing efficiency.
- Cisco developed and marketed its Application Centric Infrastructure — a proprietary policy “add-on” to routing that is often used with iWAN technology.
- VMware, through its acquisition of Nicira, has an entire family of networking products based on VxLAN tunneling called NSX. It’s standardized but contains proprietary control mechanisms that prevent standards-based interworking.
Go ahead and throw Apple and Netflix on the pile too.
We’ve established that the big guys do not talk to each other anymore. So, what about the technology startups? Today, there are over 40 SDN/NFV startups building networking technologies. Sadly, to my knowledge, none of these companies have any new networking features that will internetwork with each other, or any of the giants listed above.
With all this, it is fair to ask if we are being a selfish industry? Where have the founding principles of the Internet gone? Will our industry have the strength and fortitude to put our one-sided short term needs aside and revitalize the Internet with an infusion of new standards and capabilities? I would argue that any bright engineer would quickly say it is certainly possible to fix the Internet. At 128 Technology, we agree and believe that to do so, a few things are required:
- All private and public IPv4/IPv6 networks should be joined into a single mega-Internet with end-to-end policy routing and security — allowing for transparency and directional routing at the session layer.
- Distribution of routes for services with policies and protocols built in — allowing for control over destination addresses/ranges and access policies (i.e., IP address mask, AS#, VLAN, geolocation, etc.).
- Distribution of network topology information both detailed (local) and summarized (remote or path based) — and includes current congestion, latency, packet loss, and utilization.
Finally, we need smart software-based routers to help IT managers easily perform and manage these tasks. The technology needs to be affordable and open for all participants to access, use, and extend.
We will never reach agreement on methods and protocols, and we know commercially there will always be winners and losers. But if we can start thinking more holistically, we can begin to advance the science and art of routing.
— A Selfless, Deterministic Networker