A hub and spoke structure is one that works across multiple industries, with the airline industry being the most obvious one that utilizes this model to more efficiently take passengers from one destination to another around the world. Similarly, when it comes to the internet and digital infrastructure, data center strategists have used this method often to distribute data caching and processing across multiple geographic locations. The idea originated from some of the earliest network topologies and led to optimization of early enterprise buildouts for large financial institutions, manufacturing firms, and many more organizations with heavy data processing needs. One main hub is maintained (in decades past, this was often adjacent to the corporate headquarters) for the heaviest processing and analysis, with other smaller facilities located at regional locations or closer to end users.
As data centers moved from enterprise control to specialized operators, the transition led to several core primary markets [Northern Virginia or Chicago in the United States, FLAP (Frankfurt, London, Amsterdam, Paris) markets in Europe, Singapore or Tokyo in Asia] becoming centralized hubs for all data center functions. With the advent of an increasing number of “metaverse” applications requiring compute as close as possible to the end user, the edge spokes have and will continue to appear in increasing frequency in coming years.
Examples of hub and spoke at the market level have abounded in countries with large population hubs that are far from coastlines and thus undersea cables. These cities require backhaul fiber linking the coastal city to the major inland market, with certain processing and peering functions occurring locally at the port, reducing latency as information is sent across the ocean. France has grown into this model, with Paris a longtime primary European market and Marseille an increasingly intriguing coastal hub. Eight submarine cables call Marseille home, with another five coming online shortly mainly linking the city to the increasingly important Africa and Middle East regions. South Korea functions similarly, with the six cables (and one upcoming) entering Busan at the far southeast tip of the peninsula requiring connectivity to the global megacity of Seoul. Madrid is the centrally located capital of Spain, with Barcelona of increasing interest as a coastal data center hub with upcoming links to similar destinations as Marseille.
While many content streaming, gaming, and potential virtual reality applications are focused on the spokes at the edge, optimization heading the other direction will likely continue as workloads that do not require user proximity will be moved to the most cost-effective locales. This would include research and development of new technologies such as artificial intelligence, simulations of aircraft and automobile performance (also likely using A.I.), heavy data modeling and analysis, and other “back office” compute needs. Locations such as Quincy, Washington, Montreal and Quebec City in Quebec, and even major markets with relatively low-cost renewable energy such as Portland are ideal locations to save on operational costs and boost green credentials.
With a continued focus on optimization to maximize performance and energy efficiency, expect further development of the hub and spoke architecture for hyperscalers, enterprises, and the data center operators that serve them in coming years, with diversified requirements across main hubs and edge locations. An array of secondary markets will scale accordingly, with increasing rack density for specialized needs close to the user and more general compute needs moved away. Distributed cloud thus becomes a reality, with unified application support in all locations. A diverse, multifaceted data center ecosystem awaits!
Author: Kevin Imboden