The carrier-neutral interconnection and data centre market has been consistently growing year over year. This represents the confluence of several trends, such as a secular trend to outsourcing as companies refocus their energies on achieving competitive advantage based on their core business and core competencies, and thus choose to outsource what Geoffrey Moore calls "context" activities-such as data centre operations-to firms that can execute better based on their strategic focus.
Pragmatic constraints are yet another reason for selecting a colocation provider: enterprises often run out of space in their current data centres due to business growth that is more rapid than usual, or need to recover operations after some sort of disaster-say, flood, fire, or hurricane-and can't afford delays to select a location and build and outfit a new facility.
However, in addition to all of these reasons, there are also mathematically quantifiable benefits to siting computing and storage resources in interconnection/colocation facilities. One obvious one is the difference in the number of interconnections required on a pairwise basis rather than through some sort of facility or exchange. If n networks need to interconnect, on the order of n2 interconnections would be required on a full pairwise interconnection basis, but only n connections are needed to a meet-me room or exchange.
Another interesting benefit arises when computing and storage are sited at a colocation facility which is also an interconnection facility which interconnects those n networks. As the accompanying paper shows, such a facility can often be the best location to site such services when they are highly interactive. A number of case studies from large search providers and online shopping providers show that user experience can correlate directly to revenue. Slow sites reduce revenue as customers decide to stop shopping or shop elsewhere. Even milliseconds of latency can make a difference.
Although an exact characterization of latency depends on various factors such as the location of users, their network providers, network provider topology, and interconnections and inter-Autonomous System routing, the general problem can be modeled abstractly.
In such a model, a user on one network can efficiently reach "local" services deployed at nodes on that same network. However, for a user on one network to access a service on a different network can require traversing multiple network links and interconnection facilities to get to a service deployed at a central node on the second network, and even more links to reach a service deployed on an edge of that network, say, in an enterprise data centre.
To put it another way, in a world of many networks, some users may experience excellent response times when the service is on the same network, but a vast majority of users will not. Consequently, a central location can offer the best average performance, and such a central location can often be an interconnection and colocation facility. In one pro forma analysis, interconnecting multiple metro or regional networks through an interconnection facility could reduce response times by half, driving improvements to application performance and user experience.
In short, there are many reasons to select an densely-connected colocation provider, including firm strategic focus on core instead of context, capacity planning, total cost reduction, and user experience.
Visit www.telx.com/interconnection to view the Telx Interconnection Analysis White Paper.