The Value of Meshiness
The Benefits of Dim Fiber Services
Whenever I walk into one of my daughter’s rooms, I always think “wow, this is really meshy.” In the world of networking, the more meshy a network is, the closer it is to “full connectivity between every pair of devices,” the more complex it is. The complexity that comes with a meshy network, however, has a tradeoff: meshy networks tend to provide more optimal paths, and hence tend to support higher performance applications.
The Internet, however, seems to be losing its meshiness.
According to at least one source, the delay between two locations within the same city in China can reach as high as one second. Given the distance within the bounds of a single metropolitan area is relatively short, which means the optical time delay across the geographic area should be small, why should there be such high delays? A hint can be found here: the traffic is passing between the two largest providers in China. These two providers are strong competitors, which gives them incentives to keep as much traffic within their respective networks as possible.
One way to capture traffic is to encourage customers who already use your services to bring their entire “network buy” onto your network, and one way to encourage a company to use only your provider is to control the usability of connections purchased from other providers. Under provisioning the links between the two providers appears to give both providers a stronger bargaining position with their customers, so both providers do, in fact, under provision their links with one another. Because of this under provisioning, there is probably more bandwidth between China and local foreign nations than between the two providers.
China is not unique in this regard; much of Latin America faces precisely the same problems. Latin America, which contains smaller nations who are not always friendly (politically) towards one another, has a similar situation. The fastest path between any two points in Latin America is far too often through Miami, the first point at which multiple providers serving different geographic regions meet.
This problem is not just related to one or two geographic areas, like China or Latin America, however. The shifting shape of the Internet is pushing content towards a small set of destinations, rather than towards a widely dispersed set of web sites and services. According to a 2016 report, traffic from a small group of social media and video providers represent more than 70% of the traffic on the Internet.
Traditionally, the Internet is rather meshy; most systems connected to the Internet can be reached in about 4 AS hops, and the speed of light is often a primary determining factor in the round-trip time between two devices. A more hub-and-spoke shape will likely decrease the time to reach a specific set of services, while increasing the time required to reach all the rest of the services available on the Internet.
Will this movement away from strong meshiness at the edge and in the core continue? The move towards a less meshy, more hub-and-spoke style of topology will probably continue for the foreseeable future, particularly as businesses move towards cloud based services, and consumers move from laptops as their primary “internet consumption device” to mobile devices, such as phones and tablets. As data is centralized, traffic—and the capacity to support that traffic—will follow.
Should providers and policy makers—and perhaps even edge network operators—be concerned about this shift? There are several reasons they should, perhaps, be concerned. Consider this: once the traffic moves towards a small set of centralized data points, it may be difficult to rebuild the meshy sort of connectivity the ‘net seems to thrive on today. Building out connectivity beyond a geographical region requires some form of commercial driver producing a solid revenue stream. If the ‘net shifts too far in the centralized direction, it may be difficult to find a critical mass of such drivers in the future. Centralized data, and hub-and-spoke data flow, can potentially be damaging to large, creative segments of development, such as blockchain technologies and fog computing.
What can the average network operator do? First, ask more questions about the shape and path of circuits within and outside regions, especially when working with connectivity that runs between providers. Make certain you understand the interprovider connectivity model, where exchange points are, and how the shape of the local connection will impact your connectivity. Second, it is worth considering not jumping to the “largest” or “default” provider for cloud services, and it is always worth considering the correct balance of placing data into a cloud service, and keeping data in insourced private cloud services.
Agility—the ability to move data and processing loads between different services, and the ability to shape business processes around available connectivity—is always the key to dealing with uncertain times. Awareness of how the ‘net is changing, and what it could mean for your business, is always the first step.
For more information on inter-carrier connectivity, download our latest whitepaper.