Can you choose your carrier when a carrier owns the data centre?

Twenty years ago when I began using hosted data centre services, the choice was straightforward. You picked either a co-location service where you rented rack space and installed your own servers, or a hosted option where your data sat on the service provider's equipment.

You could then decide on the amount of internet bandwidth you needed, and there was often an option to pay an additional fee for a resilient link to provide better uptime guarantees.

Today, the story is very different. You can still do the above if you so desire, but in many setups you have an extra option: you can choose the carrier(s) for the connectivity between your installation and the outside world.

(This generally means internet connectivity, but it could equally mean direct private WAN or MAN connections if your company is big enough to afford it.)

For example, a UK hosting provider that was among the first carrier-neutral hosting providers I came across many years ago now has 54 carriers with a presence in its various data centres.

However, that is a company that simply owns data centres. It is not a telco or an internet service provider: it just rents you the space, keep the lights on and the environment consistent, and provides you with an interconnect between your kit and the comms area in which the carriers have their internet presentations.

Maintaining neutrality

But what if the owner of the data centre is a telco as well? Is it possible for it to be truly independent and neutral when it comes to connectivity you can sign up to?

To have a choice of network vendor in any data centre one of two things has to happen. The first is that each of a number of vendors has a permanent point of presence (PoP) in the data centre, and the data centre owner provides an inexpensive cross-connect service between the PoP and your cabinets

The second is that the data centre owner must be willing to provide wayleave permission for whichever carrier you nominate to route its connectivity into the building and into your area (again the internal component could be a cross-connect provided by the data centre owner).

The first of these options is by far the most attractive to the carriers – a single PoP can service dozens of customers – and it is probably just as attractive to the data centre owner too because digging up the road and routing cables through entry points carries a risk of disturbing other nearby services.

Traffic flows

The telco market follows the reseller model: the telcos route traffic over each other's physical networks and in many cases simply re-sell each other's services.

So if you rent a connection from a given provider, the chances are that the traffic is routed over networks belonging to a variety of other telcos.

This is, incidentally, a Very Good Thing. The product of the vendors routing via each other's worlds is a genuinely resilient network: if one vendor has a problem, the traffic will route via a different path over other vendors' links.

At a local level, though, life can get a bit interesting for a couple of reasons.

The first is that in the average large city there is usually one primary telco, or sometimes two. There may also be some smaller companies providing services to the most densely populated areas via their own networks, but it is uneconomical for them to have their own cabling anywhere else. They fill the gap by renting services on a wholesale basis from the larger incumbents.

So if you are looking to host services in a data centre owned by the primary local telco, it may be the only company with cabling into its own premises. Even if you rent connectivity from another telco, the last mile may well belong to the telco that owns the data centre anyway.

Behind the scenes

The second aspect to consider with local connectivity is that in places the carrier world mimics the approach used by power companies in many countries.

In the deregulated power industry you can buy your electricity or gas from one of a dozen providers, but if you switch provider you don't get a guy in a boiler suit knocking on your door and installing a new connection or a new gas pipe.

You are using the same electricity or gas as your next-door neighbour. You may pay different companies for the service, but behind the scenes they are just settling up between them via reseller agreements they have with each other.

The same can be true with carrier services: in telco A's data centre you might buy connectivity from telco B which is actually provided entirely by telco A via a white-label agreement.

So, can you get carrier-independent services in a carrier's data centre? It is not sounding promising thus far: we have competitor services that are white-label versions of the data centre owner's services, and carrier-owned data centres serviced only by the carrier's own connections.

But it is not the end of the world. Where the data centre owner owns the only physical connectivity into the premises, it is generally the local monopoly and its regulator will force it to offer a competitive wholesale product to competitors.

In turn the competitors will use as little as possible of the data centre owner's services (either dark fibre or layer 2 – simple, point-to-point services), so the vast majority of what you buy will be your chosen carrier's.

More importantly, where the data centre owner permits other carriers to have a presence in its data centre it is acknowledging the key fact that carrier services are a commodity.

The owner is probably paid a rent or a commission by each of the carriers, and this may well bring at least as much profit as it would gain from providing the network services itself.

Let’s face it, managing ISP-scale carrier services is not a simple matter. It requires expensive infrastructure kit and support staff. And while I remember the days when data centre owners would charge the earth for a 50-metre cross-connect between cabinet and third-party comms rack, you seldom see that now.

In fact there are only really two gotchas with carrier services in the data centre, and neither is really down to anything the data centre owner does.

You could find out the hard way that one vendor is just reselling a service from the other

The first problem comes where one carrier uses another's local loop. If you have, say, a four-hour response agreement with your carrier, it may well have a two-hour agreement with its upstream provider and its contract with you allows it to add the two together.

The solution is easy: negotiate a better agreement. It is perfectly possible; I have done it more than once.

The second problem comes when you decide to use two different carriers in the interests of resilience. As well as having the fun of getting two vendors to work together to provide intelligent failover between their services, you could find out the hard way that one vendor is actually just reselling a service from the other so you have gained nothing by purchasing from them both.

The answer is simply to ask your carriers to explain any third-party contributions to the service provision so you can spot potential issues.

Too big to fail

Data centre providers make money by renting you space and guaranteeing that the lights will stay on and the room will maintain a constant temperature, the right humidity and a lack of smoke or flames.

But when it comes to the bandwidth between them and the world outside, there is money to be made by hosting third-party connectivity on-premise as well as selling one's own bandwidth to customers.

Additionally, regulators make sure that carriers' wholesale relationships remain competitive and that faults are serviced promptly for their reseller customers, even when they are a near monopoly (or especially when that's the case).

And many carriers are content to live without dishing out a few megabits to each of a few hundred piddly little clients. It is a commodity industry where it makes sense to let others take the burden and support hassle.

So yes, it is perfectly possible for a carrier-owned data centre to be carrier-independent. ®