Om Malik has a story at Business 2.0 speculating on some of the reasons Google have been alleged to be buying high capacity dark fiber capacity from wholesale telecommunications providers. This comes several months after CNet published a speculative story about why Google were advertising for someone with experience negotiating with telecommunications companies. The complete list of job qualifications for said listing are as follows:
- Negotiations for collocation space in conventional data centers; for racks, power circuits, cross connects, and remote hands services in conventional data centers; and for wholesale transactions with conventional data centers in North America, Europe, Asia, and Latin America.
- Negotiation and purchasing of IP transit services in North America, Europe, Asia, and/or Latin America; negotiation of partnerships with Internet exchanges, regional peering providers, and paid peering arrangements with major carriers.
- Identification, selection, and negotiation of dark fiber contracts both in metropolitan areas and over long distances as part of development of a global backbone network; contracts and negotiation for managed metropolitan services and long haul wavelength services to fulfill capacity and redundancy requirements in North America, Latin America, Asia, and Europe.
- Identification and negotiation of contracts related to leases or purchases of data centers facilities and/or properties capable of conversion to data center purposes; experience with evaluating and assessing potential data center facilities for acquisition; experience negotiating startup, service, and maintenance contracts for data centers; experience obtaining data center infrastructure hardware including chillers, generators, UPS systems, transformers, power distribution units, etc.
My own telecom experience didn’t lead me to think too, too much of this at the time. The original job posting looks pretty standard for a company with heavy data hosting requirements that is going to need lots of hosting space and bandwidth capacity to support a data-intensive platform such as Google. Considering the volumes of information they’re dealing with, dark fiber seems like a perfectly cost-effective path to take. The per-unit cost of dark fiber is significantly less than leased capacity, but only the highest-volume customers can make it cost effective. Likewise, Google put so much information onto the Internet backbone, that they must pay a lot in interconnect charges. As Google begin to offer more and more high-bandwidth content such as audio and video, it would become a significant expense to their bottom line. Using dark fiber and effectively building their own Internet backbone would help them better control these costs.
Om speculates that one possibility is that Goolgle could build out a nationwide Wi-Fi network. This is the bit of his article which is likely to get the most attention. The business model would be a mix of subscription access and location-based advertising. I’m not sure dark fiber purchases lend themselves to this end, because there would be an awful lot of local access connectivity to individual Wi-Fi locations which would be required. However, Om may have better insider information than I: Business 2.0 has learned from telecom insiders that Google is already building such a network, though ostensibly for many reasons.
The Wi-Fi thing doesnt make sense to me. Wi-Max maybe. But I doubt Google would want to deal with all of the local access issues involved in a major Wi-Fi build out (dark fiber wont help you here) nor the negotiations with owners of the venues in which Wi-Fi would be made available.
A reason which makes much more sense to me, and that Om outlays in his article as well, is that taking a dark fiber approach helps Google eliminate the expense of putting large amounts of information onto ISP networks. Om writes:
An even more compelling reason for Google to build its own network is that it could save the company millions of dollars a month. Here’s why: Every time a user performs a search on Google, the data is transmitted over a network owned by an ISP — say, Comcast (CMCSK) — which links up with Google’s servers via a wholesaler like AboveNet. When AboveNet bridges that gap between Google and Comcast, Google has to pay as much as $60 per megabit in IP transit fees. As Google adds bandwidth-intensive services, those costs will increase. Big networks owned by the likes of AT&T (T) get around transit fees by striking “peering” arrangements, in which the networks swap traffic and no money is exchanged. By cutting out middlemen like AboveNet, Google could share traffic directly with ISPs to avoid fees.
This makes much more sense to me, and fits in better with Google’s current content-oriented strategy.
