Mike Hammett
May 19, 2017

There was some commentary at the end of yesterday’s CHI-NOG (Chicago Network Operators Group) meeting that to quote an attendee outside of the peering community, “was a huge disservice to attendees that didn’t know any better”. The commentary was that “peering is dead” and that fostering ecosystems and interconnection is where we are and should be going. They continued that the importance of these carrier hotels wasn’t the Internet exchanges or the peering that took place, but the driving down of transit prices through increased transit competition. The Internet exchanges outside of the major markets were just community “good will” from large corporations and had an insignificant impact on performance. We don’t disagree that it’s about ecosystems where you can do things you couldn’t do before. We don’t disagree that the traditional peering models of 20 – 30 years ago are dead. We don’t disagree that competition is good and has been driving down prices in all markets. We don’t disagree that the major CDNs play a huge roll in Internet exchanges. However, if you’re new to what peering is, you don’t know the context of the comments. You’re lead to believe that these Internet exchanges are just peddling snake oil.

Peering is still very much happening and very much important. People like Midwest-IX, RVA-IX, NashIX, MGMix, etc. are bringing Internet exchanges and “peering” to people and markets where there was no established place for such activities before. Hundreds of local networks are becoming aware of the advantages of peering, be it cost, performance or additional capabilities. The community in general talks about the value in new Internet exchanges being the “unique ASNs” they bring to the table. It could be universities or other educational networks. It could be small mom and pop last mile providers, data centers, hosting companies, managed services providers, IPTV or VoIP providers. Through these Internet exchanges and the peering they facilitate, traffic is kept local. There is no longer a reason for traffic within Indiana or Missouri\southern Illinois to go to Chicago or traffic in Alabama to go to Atlanta or Dallas or both. Someone on a local ISP can now connect directly to their local university for online classes, accessing research or simply working from home without fear of an unpleasant experience due to delay and congestion.

Big content and the tech media talk about pushing content to “the edge”. That’s exactly what these Internet exchanges are doing by building out in tier 2 and 3 markets. They’re providing fabrics closer to the end-users to reduce the cost of service delivery as well as increasing the performance of said delivery. Many of these networks are already in the facilities the exchange would be in or the data center would partner with the IX to bring the IX to their facility at the risk of becoming irrelevant in the marketplace. The cost to join at that point would vary depending on that data center’s cross connect model. Fortunately, many data centers in these markets have very operator-friendly cross connects. Regarding the performance, we continually hear from networks that join our exchanges that overall traffic increased once they joined. It could be the effects of latency on TCP performance or it could be that there was congestion above them that they’ve now managed to circumvent. There’s not a lot of opportunity for something to go wrong when source and destination are connected locally through a high-performance switched network.

The Internet has changed significantly in the past 20 – 30 years when networks were a lot more hierarchical. End user made a request that went to their ISP. It may have traveled to another ISP before hitting a “tier 1” provider and went through their peering to another “tier 1” before descending down the other side to the other “tier 2” provider or perhaps their ISP did peer directly with the other “tier 2” ISP, whether direct or through an IX. There were few to no central repositories of content. Content was everywhere and that very much supported that hierarchical system. You depended on other networks to take you to other networks because there were so many places to get content from. Today there are still as many and perhaps more places to get content from. However, traffic profiles have changed. Now a single network is the source of 1/3 of all Internet traffic, Netflix. Another third comes in aggregate from a few other sources like Akamai, Google, Microsoft, etc. An ISP that largely caters to residential customers can move as much as 85% of their traffic over to Internet exchanges. When so much of your traffic comes from so few places, the hierarchical model simply doesn’t make any sense. While not being a peer from the historical or strictest sense, an end-user ISP and big CDNs are highly complementary. It makes incredible sense to directly connect these networks together.

But aren’t most ISPs trying to inhibit usage of these streaming video sources? Don’t these big content networks already have agreements with the ISPs? That is true of many in the top 10 – 20 (maybe even more than that) ISPs, but there are literally thousands of ISPs in the United States, not to mention higher education institutions that have more users than many ISPs. The general public decries the power the largest ISPs have, yet few (consumer or technology\content leaders) are willing to give the other ISPs the tools necessary to be successful. Marginalizing them isn’t going to do anyone any good. Other than the obvious patronage of customers, these other ISPs need the tools to serve the content the customers want cost effectively. These tools may be Internet exchanges, they may be caching boxes, they may be something else entirely. However, Internet exchanges and “peering” have had a big impact on many of these networks and has helped them to continue to provide competitive services to hungry customers.

Peering isn’t dead, it’s just different and there’s nothing wrong with that.

Previous CHI-NOG peering sessions for those looking to learn more: