You write about the value of IXPs... what are the forces that cause their formation, and what causes their demise?
Consultant, Initech Corp.
Competition encourages us to stretch and innovate, to create new products that delight the market, that abstract complexity into simplicity, and carefully control the flow of information. But there are few forces leading competitors to cooperate.
One thing that makes Internet Exchange Points (IXPs) so interesting is that the aligned interests of competing network operators drive cooperation, and that the cooperation is implemented in the form of the birth of an IXP.
Cooperation encourages us to share enough information in order to identify aligned interests, and potentially act together, at least until those aligned interests aren’t strong enough to support continued coordinated pressure in that direction.
Consider the alignment pattern that created IXPs in Europe in the 1990‘s. A collection of ISPs sent their Internet traffic through the US for offloading to the other Internet network operators there. Much of this traffic is ultimately delivered to their competitors in their home market. But not only does this transport to the U.S. cost money for both parties, but there is also high latency. End-user performance is bad since the traffic has to traverse the ocean twice, and the cost is so high that few are attracted to this sub optimal Internet thing.
As shown below, these pain points exceed what I call “the IXP Formation threshold” leading ISPs to share information with one another to determine that they should exchange at least their local traffic directly and locally with their competitors, and do so by forming an organization to operate that infrastructure to their common shared and aligned interest.
These aligned interests are very narrowly defined as the direct exchange of traffic among themselves within the boundaries of the stage of cooperation, with “insolation” filling the gap between the competing and cooperating interests vehicles as shown below.
So for these narrow activities, the gravitational pull of the new IXP supports the network operators dual competition and cooperation stance.
Strength of Cooperation
So how strong is this motivation to continue to cooperate?
Based on the interest of members at IXP member meetings (as calculated by the percentage of members actually voting) the interests are sufficiently aligned to maintain continued operation of the shared infrastructure, but not enough to dramatically change course with any assurance that the members’ collective best interests are at heart.
The community brought these IXPs forward because they served a narrow aligned interest; in the absence of strong voter turnout, all other pursuits must be assumed to be for the employment interests of the IXP operations staff themselves.
Three Types of IXP
This leads to three distinct types of IXPs in the market as shown in the figure below.
The High Value-Add IXPs
The High Value-Add IXPs continue to add projects and activities into their mix of responsibilities beyond running the IXP peering infrastructure. These services run the gamut from global data collection to operating shared video services, to evangelizing peering in foreign countries and even pursuing partnerships in other ecosystems. There is no end to the list of projects that meet the “not a bad idea” bar.
So directly or indirectly, all members pay for these services. What are the IXP member dissenters to do? Unless the project is diametrically opposed to the vital interests of a company, probably nothing. And an IXP would never push projects that would irritate the wrong people. So with extra revenue and so many interesting projects being in the “not a bad idea” category, then one can understand why IXP scope will creep.
The pain in general would have to be very to pull out of an IXP. Like the “Roach Motel” as Ellen Hancock (former CEO of Exodus) called them - network operators check in but they never check out.
The Professional IXP
The Professional IXPs are IXPs singularly focused on only providing the most professional and most cost effective peering infrastructure possible. They have a complete complement of staff - administration, business development, marketing, sales, legal, etc. but they do not take on additional projects. Any excess revenue is returned to the community in the form of lower costs.
The Minimum IXP
Bill Woodcock once advocated the “IXP in every broom closet” approach. What happens if the broom closet loses power? Other peering points and Internet Transit serve as back ups. The point here is to minimize the cost and maximize the places where traffic can be offloaded across cheap ethernet switches. These minimum cost IXPs are supported by volunteers with pagers, servicing switches contributed to the community.
You will find each category has their vocal advocates and dissenters.
The IXP Failure Scenario
If the gravitational pull of IXPs is weaker than the pain of working with the IXP, the IXP is on their way to failure. The IXP is “arrogant”, “unresponsive”, “too expensive”, “difficult to work with”, etc. generally out of touch of the pain points of dealing with them. This will lead the ISPs to align against a common enemy (the IXP) and seek alternatives.
In a nutshell, competition may delivery suboptimal outcomes when it ignores the broader benefits of growing an entire ecosystem of industry players working towards aligned interests. It is difficult but it is possible, and IXP formation is a wonderful thing to see. It is the starting point to reduce the cost of Internet traffic exchange. And when the IXP works, it gains momentum, it becomes extremely powerful in the Internet Peering Ecosystem, and becomes extremely difficult to compete against.