What makes a Colo?

Ask Dr. Peering

Q:
DrPeering -

In a bind here. For our new Pied Piper compression delivery next week, we need racks to host our custom hardware, and colocation with big networks. Transit, Peering, Remote Peering, and a blend of CDN and Cloud connectivity I figure. Any suggestions?
-- Bertram Gilfoyle
A:
Gilfoyle -

I hear your stress, but not to worry - Urgency is my middle name.

And it turns out that this is a very common story for high performance Internet applications such as high rate compression of video streams. I would start with a focus on markets (see last month’s article) and then the careful selection of colocation. Go for a colo with the right blend, which brings me to my research. Permit me a brief tangent here...

I’ve been doing a lot of modeling of the Internet Peering Ecosystem, using Swift as a language to describe the components and the messages between the elements. My research has been validated with a number of phone calls these days with network operators, content providers, enterprises, and increasingly with Tier 2 market colocation centers. This process of modeling has been challenging some of my underlying assumptions.

For example, we all think of a colocation center as a purpose built data center for colocating different networks, usually for peering. Through colocation, networks can easily and cost effectively interconnect with one another, reduce their dependance on their transit providers, etc. and the value of the colocation center is proportional to the value of the colocated networks. (See below.)

Now consider a colocation center with *no peering*... is this STILL a colocation center? (See below.)

The value is still proportional to the value of the population from the perspective of a network operator. There may be tremendous value having all of these transit providers available in one location. Even though the routes aren’t provided on a settlement-free basis, as is typically the case with peering, there is still value in having direct access to these transit providers.

The problem I have run into during my consultations is that some colocation operators proudly convey their ignorance of their population. “We don’t get into their business” they proudly claim, “we are neutral.”

This makes me crazy.

The problem is that these guys do nothing to facilitate the interconnection between their customers, and therefore they are missing the greatest opportunity to increase stickiness. The more interconnections in the building, the greater the value derived from being in the building, and the more painful it will be for them to leave.

The best colocation centers curate their collection of high-value participants, and facilitate a hybrid of interconnections as shown below. They will understand the business of every participant, have a personal friendship with each of their operators, and encourage the cross pollination of ideas through parties, peering workshops and lectures, book signings, etc. They will bring forward the next Nike of Internet services, one that will combine the best of breed off-the-shelf network services into a new service, expertly marketed and delivered.

Colocation providers should know their customers. This is ultimately their value proposition. If you don’t get this from your colocation center, find another, one that is more likely to grow its population. When the population grows, so too does the value that you will derive over your lifetime of working with them.

I hope this helps -

After modeling and validating my model in the community, I stumble upon my own decades of assumptions. Assumptions should be questioned. And so, here is one assumption I was making that deserves to be ripped apart.

What makes a Colo?

May 31, 2015

The 2014 Internet Peering Playbook

In Print

and for the Kindle:

The PDF, ePub and

.mobi files are

also available as a

perpetually updated

DropBox share:

Price: $9.99

and in French!