Exaflood. Or not?

The current economic environment has raised many questions about what the future holds for the telecommunications and Internet industries. While no one disputes that Internet traffic continues to grow in spite of — or perhaps because of — the macroeconomic downturn, many have questioned whether service providers will continue to invest in their networks, or whether they will simply “run their networks hotter.” By running a network hot, service providers try to squeeze more performance out of their existing assets, delaying capacity upgrades until absolutely necessary. This strategy can certainly save costs in the short term, but it is not without risk, as service quality and reliability ─ especially in times of congestion – will be impacted.

In all likelihood, some service providers will run their networks hot in some places —after all, everyone is trying to do more with less. But there are some networks and services for which running hot simply isn’t an option. However, before we explore the phenomenon of running networks hotter, let us take a look at current network utilization levels. In a follow-up blog, I will explore the myths and realities of running networks hotter; where it makes sense and where it doesn’t.

Andrew Odlyzko (University of Minnesota), one of the nation's top experts on global Internet traffic, notes that while network traffic growth is strong, it doesn't necessitate upgrades. In fact, he states that 2008 was the second consecutive year that network capacity grew faster than traffic, indicating that utilization levels are going down. New data from TeleGeography's Global Internet Geography draws the same conclusion, stating that between 2007 and 2008, average utilization levels decreased from 31 percent to 29 percent, while peak utilization fell from 44 percent to 43 percent. However, the overall trend toward lower network utilization levels belies significant regional differences. While utilization on international links to Europe and Asia fell in 2008, traffic growth outpaced the deployment of new Internet bandwidth in the United States, Canada and Latin America.

These reports are contradictory to the stories of future floods of video wreaking havoc on the Internet. They are also inconsistent with the feedback we are receiving from our customers. So, why the difference?

As Odlyzko mentions on his Web site, he uses publicly available traffic statistics that are not a representative sample of all traffic on the Internet. The 100 or so sites he monitors are mostly public Internet exchanges and academic networks, and include very few private companies or commercial service providers. This means his data doesn’t take into account private peering, which is by far the most common method of interconnection for the Internet. Hence, he warns, great care should be taken in drawing any conclusions about the overall state of the Internet from the analysis and data on his site.

Furthermore, simply looking at historical data to model future traffic growth is fine in a deterministic environment (like the stable TDM networks in the past). But what if something unpredictable happens? Statistical models can’t project when the knee in a bandwidth utilization curve may happen, just like they couldn’t predict the events leading up to the current economic crisis.

I believe we are right at the beginning of the unpredictable; we are witnessing yet another behavioral shift. Just like the music industry had to transform itself after the introduction of the MP3 player, the broadcasting industry will have to reinvent itself to survive the coming wave of Internet-based entertainment. Technology such as AppleTV, Sling, Roku, Xbox, PS3, or the new Blockbuster set-top box) are likely to cause a dramatic shift in consumer viewing behavior. For the skeptics, just look what the iPod did to the music industry. If your family is anything like mine, TV viewing time is going down rapidly in favor of using computers and the Internet for entertainment. And this behavioral shift will cause a surge in traffic that could stress the networks to the breaking point. So far, these traffic spikes were caused by individual events such as the Super Bowl, the US Open golf tournament, or the Obama inauguration, which caused clear failures and traffic drops in at least two networks during the traffic peaks. Just imagine what will happen if primetime high-definition TV (multicast) and video-on-demand (unicast) are delivered over these networks.

Yet another interesting angle in this debate is the impact of traffic growth on the access network versus the core network. The access network is really where the bottleneck is. Core networks use advances in DWDM technology to continuously increase the capacity carried over a pair of fibers. Links that previously supported eight or 16 channels at 2.5gbps, can now carry 64 or more channels at 10 or even 40gbps, with 100gbps per channel just around the corner. In the access network, however, the endpoint of each line has to be upgraded individually: Higher bandwidth on a cable network may require a DOCSIS 3.0 cable modem; higher speed on a copper loop may require VDSL which can serve only shorter distances; FTTH carries steep deployment costs. One area where there are definite signs of rapid data traffic growth is wireless. Accommodating the growing number of high-speed broadband mobile devices requires upgrading the radio access network, which is limited not only by the number of connections it can support, but also by the total amount of bandwidth that it can deliver (see my earlier blog on the Mobile (R)evolution).

The above examples show that networks are straining to accommodate the growing traffic. While the access network seems to be the biggest bottleneck, our customers are telling us they need to keep adding bandwidth to the core of the network as well. In fact, they have been doing so since the Internet became mainstream 10 years ago. Yet, the Internet was always able to adapt and survive. Throughout the last decade it continuously transformed itself to support the flood of new applications and traffic it had to carry. In the early days, the network strained to support images, then animated images, followed by flash video and more recently streaming video. Along the way, the number of users kept growing, families added more computers and worldwide adoption continued at a rapid pace. The addition of high-definition video is the latest challenge for this relatively young medium, but we can be pretty sure it won’t be the last. It will survive; have faith in the broadband revolution!

Luc Ceuppens is vice president of product marketing, High-End Systems Business unit at Juniper Networks(JNPR).