Saturday, February 15, 2014

Asymmetrical Traffic Dominates Networks; Will Affect Interconnection Wars

Asymmetrical traffic demand has been a technology and business issue in the networks business for some time, and is becoming an issue again as video traffic starts to dominate all traffic types, and networks have to be fundamentally redesigned to account for the new traffic characteristics.

The biggest single change is that IP network traffic increasingly is dominated by highly-asymmetrical entertainment video.

Metro traffic will surpass long-haul traffic in 2014, for example, and will account for 58 percent of total IP traffic by 2017.

Metro network traffic will grow nearly twice as fast as long-haul traffic from 2012 to 2017, propelled by the increasingly significant role of content delivery networks, which bypass long-haul links and deliver traffic to metro and regional backbones.

Likewise, content delivery networks will carry 51 percent of Internet traffic in 2017, up from 34 percent in 2012.

Globally, IP video traffic will be 73 percent of all IP traffic (both business and consumer) by 2017, up from 60 percent in 2012. The sum of all forms of video (TV, video on demand [VoD], Internet, and P2P) will continue to be in the range of 80 and 90 percent of global consumer traffic by 2017.

But significant business issues are raised, not just technology and architectural issues.

Though some of the implications mostly are about efficiency, many of the traffic-related issues have revenue and business implications.

In years past, the Federal Communications Commission has had to consider situations where asymmetrical traffic flows have cost implications for service providers that are one-sided and distort the normal economics of traffic exchange.

Historically, the assumption was that traffic would be symmetrical, both inbound and outbound on any single network, and then symmetrical across network boundaries. That had obvious implications for network design and also the arrangements whereby networks compensated each other for terminating or carrying traffic.

But business relationships between networks have been, and will be, affected by various types of arbitrage, especially when traffic flows or retail rates are asymmetrical.

Traffic pumping, also known as access stimulation, is a controversial practice by which some local exchange telephone carriers in rural areas of the United States inflate the volume of incoming calls to their networks, and profit from greatly increased intercarrier compensation fees.

They do so by operating inbound toll-free calling services services where traffic is, by definition, unbalanced. Back in the days when dial-up Internet access was the norm, the same sort of arbitrage existed.

Modem pools could be located where favorable long distance termination rates could be leveraged.

Phantom traffic is another form of interconnection arbitrage.

Granted, IP network interconnection is governed by different rules than common carrier services.

But the same potential for arbitrage can exist, if networks with highly-unequal traffic flows are forced to interconnect on a settlement-free basis. That already is an issue for interconnecting content delivery networks or networks handling video applications such as Netflix, which by definition are characterized by huge downstream flows and almost no upstream traffic.

For that reason, the potential for arbitrage will exist, if IP networks are forced to interconnect on a settlement-free basis. Proponents naturally will put the best light on such mandatory, settlement-free interconnection by arguing it is necessary to maintain lawful content access.

In other words, the mandatory settlement-free interconnection will be couched in “equal access to content” or “network neutrality” terms.

Others might say the issue is business arbitrage.

And that potential arbitrage will grow as all public IP network traffic comes to be dominated by content, specifically entertainment video.


No comments:

"Tokens" are the New "FLOPS," "MIPS" or "Gbps"

Modern computing has some virtually-universal reference metrics. For Gemini 1.5 and other large language models, tokens are a basic measure...