Q. Why has Cisco developed the Visual Networking Index - Forecast and Methodology?
A. Given the ramifications and implications of prior Internet traffic growth rates, Cisco felt compelled to provide a realistic forecast that is based on multiple levels and sources of real data and projections. While this data is obviously of great interest to Cisco, we also feel that our customers (in all segments) and the industry at large might benefit from our findings.
Q. What is visual networking?
A. The variety of consumer and business IP networking trends are largely being driven by video and social networking/collaboration (such as web 2.0 technologies), the combination of which is termed visual networking. A visual networking experience can range from a prearranged TelePresence meeting to the fluid and flawless delivery of video to any device a consumer chooses - be it a TV, PC, or mobile handset.
Q. How often is the Cisco Visual Networking Index - Forecast updated?
A. The Cisco Global IP Traffic Forecast is updated bi-annually (every six months). Updates will typically be provided in the June and December timeframes. The initial Cisco Global IP Traffic Forecast was introduced in June 2007.
Q. What is the methodology behind the Visual Networking Index - Forecast?
A. The forecast relies on analyst projections for Internet users, broadband connections, video subscribers, mobile connections, and Internet application adoption. Our trusted analyst forecasts come from Kagan, Ovum, Informa, IDC, Gartner, ABI, AMI, Screendigest, Parks Associates, and Pyramid.
Cisco also collects traffic data directly from a number of our service provider customers, and this data is used to validate and adjust the usage assumptions underlying the forecast model.
Q. Have there been any methodological changes in the latest forecast update?
A. The only methodological change since the January 2008 forecast is that the "Multinationals" category of business traffic was eliminated. Instead of treating multinationals separately, traffic from offices of multinationals is now counted in the respective regions where the offices are located.
Q. Have you changed your assumptions in the latest forecast update?
A. When the forecast is updated, all the underlying analyst forecasts are updated. Any changes in analyst forecasts will be reflected in the traffic numbers. For example, the analyst firm Cisco relies on for video subscriber forecasts has increased its forecasts for commercial video adoption in central Europe, with the number of digital subscribers once anticipated for 2015 moved up to 2012. The growth of digital cable subscribers has been even stronger than anticipated, so the analyst adjusted the numbers accordingly. In addition to updating the underlying analyst forecasts, Cisco will adjust usage assumptions in response to changing consumer behavior.
Q. The growth rate seems high. I have heard that Internet traffic growth is slowing.
A. The model estimates that Internet traffic grew 46 percent in 2007. This growth rate is in line with estimates from Andrew Odlyzko at the University of Minnesota, widely recognized as an authority on Internet traffic. Cisco does project a higher growth rate of 51 percent for 2008, followed by 44 percent for 2009 and 40 percent for 2010, and continuing to decline in subsequent years. Cisco's forecast is considered conservative by most industry analysts.
Q. The growth rate seems low, given the rapid adoption of video.
A. There is no question that video is generating a substantial amount of traffic. However, the volumes are so large that sustaining high growth rates becomes increasingly difficult. For instance, despite a declining growth rate, Cisco is projecting that global Internet traffic in 2012 will be over a thousand times greater than all the traffic traversing the entire U.S. Internet backbone in 2000.
Q. What is an exabyte? What is a zettabyte?
A. An exabyte is 1,000,000,000 gigabytes. A zettabyte is 1,000 exabytes. The chart in Figure 1 shows a number of examples of data that reaches the exabyte and zettabyte scale.
Figure 1. The Zettabyte Scale
Source: Cisco, 2008
Q. Why have the historical numbers changed since the last publication?
A. Most often, historical numbers will change when there is a change in definitions. For instance, Japan's historical numbers for Internet traffic are higher in this update than they were in the January publication, because some business Internet traffic that was previously included in a category called "multinationals" has been moved to Japan. Sometimes, historical numbers will change slightly if Cisco has acquired data on previous years that we did not have access to before. There is no simple way to get "actuals" for traffic, so every number in the forecast is modeled.
Q. How have the Cisco Visual Networking Index - Forecast results been received by the industry (analysts, press, and service providers)?
A. The initial Cisco Global IP Traffic Growth Forecast has been well received on all fronts. Since its introduction, Cisco has shared its findings directly with more than 30 analysts from 13 different firms. As part of these interactive discussions, we have applied much of the feedback from independent analysts to refine and expand our methodology (including traffic type definitions, applications/services tracked, measurement techniques, etc.). Here's a sampling of the analyst feedback we have received:
"Cisco is basing much of its IP NGN strategy on the expectation that global traffic growth will continue to grow at an astonishing rate through 2011 and beyond. This traffic growth is being fueled by applications such as mobile data traffic, business Internet, business IP WAN, consumer Internet, and IPTV/CATV applications. Based on data complied by Cisco, the combination of all traffic types is projected to grow at a 42 percent CAGR for the 2006-2011 period. Given this and other projections, service provider networks not only need to bulk up to handle the continued subscriber growth and increasing bandwidth requirements, but they will also need to reach new levels of intelligence and the ability to delivery multiple premium services in a cost-effective manner." - "Cisco Generates a New Competitive Level Set within the IP Service Edge Space," Glen Hunt/Ron Westfall, Current Analysis (3/4/08)
"Cisco developed a fairly extensive, and in our opinion well-researched, global IP traffic forecast, published in July 2007" - FTTH/PON Equipment Market Forecast, Jeff Heynen/Mark Showalter, Infonetics (3/10/08)
"There are fundamentally two approaches to constructing an Internet traffic demand model. The first approach is to try to create one or more typical user profiles that describe which applications each user is running, for how long, and how much bandwidth a typical application consumes as it's being run by a typical user...This approach has a couple of significant advantages, including granularity, defensibility, and comprehensiveness...Most likely for these reasons, this approach is favored by several consequential researchers in this space, most recently an internal research team at Cisco Systems Inc., which deployed this approach to model consumer application deployment in (Cisco, Global IP Traffic Forecast and Methodology, 2006-2011, 2007). The Cisco model assessed all known consumer applications generating significant amounts of Internet and IP traffic in the summer of 2007, and extrapolated these out through 2011. (Cisco also looked at enterprise applications, but with less granularity, which is reasonable as these tend to be more difficult to classify meaningfully into categories)." - The Internet Singularity, Delayed: Why Limits in Internet Capacity Will Stifle Innovation on the Web (Johna Til Johnson, Nemertes, 11/28/07)
Initial Cisco Global IP Traffic Growth Forecast content has also been featured or referenced in a variety of industry and business press articles since 2007, including The Wall Street Journal, The New York Times, Network World, Information Week, eWeek, The Mercury News, and many other print and online publications based in North America and abroad.
Cisco has also briefed dozens of its Tier 1 and Tier 2 global service provider customers on its IP traffic growth analysis. Feedback from this group has also been positive and many are interested in participating in the effort by anonymously sharing their traffic data. This collaboration will serve to broaden the scope and sampling of real data that Cisco can apply to future forecast updates.
Q. Service providers report that business traffic accounts for less than 15 percent of traffic on their networks. Why does Cisco estimate business traffic to be over 30 percent of Internet and WAN traffic?
A. Service providers usually limit their definition of "business traffic" to traffic from business WAN services such as IP VPN. In this case, the comparable metric in the Cisco model is business WAN as a percentage of the sum of consumer Internet, business WAN, and business Internet, which is 12 percent for 2007. Even when service providers include business Internet in the business category, it is generally only headquarter sites that subscribe to very high-capacity Internet connections. Smaller sites with broadband connections are generally grouped with "consumer" when reporting traffic.
Q. Why is VoIP traffic so low?
A. While immensely popular, VoIP is very lightweight in terms of bandwidth. However, it is an important consideration for service providers in that quality of service (QoS) is important for voice, and one strategy for improving QoS is to increase capacity so that there is always sufficient bandwidth for the speedy transport of time-sensitive voice and video traffic.
Q. Does this forecast include signaling traffic?
A. No, signaling traffic is not included. However, an estimate can be made using the standard rule that IP signaling traffic is approximately 3 percent of bearer traffic.
Q. Why is broadcast TV traffic so low in comparison to VoD traffic?
A. Broadcast traffic is low because it is a one-to-many service rather than a one-to-one service, like VoD. For each VoD request, a new stream must be served, whereas when hundreds of people tune in to the same television show, only one copy of this show needs to cross most of the network, until close to the edge where it is split and sent over each access line. In this forecast, the access-line traffic for broadcast TV is not included.
Q. What about satellite video traffic?
A. Because satellite is similar to broadcast in that it is one-to-many, the exclusion of satellite from the forecast is not expected to make any significant difference. However, direct broadcast satellite (DBS) providers are now deploying set-top services that simulate VoD by pushing the top 25 programs to the set-top boxes overnight and enabling on-demand access to less popular content through the subscriber's Internet connection. The on-demand streaming is certain to have an impact on traffic in the future, and has been factored in to the assumptions for Internet video.
Q. What about digital terrestrial video traffic?
A. Like satellite, digital terrestrial television (DTT) is a one-to-many service, so the exclusion of DTT is not expected to materially impact the accuracy of the forecast. Also similar to satellite, Pay-DTT providers may move to establish a broadband connection to the home in order to be able to offer on-demand content, Internet content, and interactive content. This scenario has not been included in this forecast, because the penetration of DTT remains low throughout the forecast period. DTT may be included in future versions of the forecast.
Q. Why isn't Internet gaming traffic more significant in 2012?
A. First-person shooter (FPS) games are the most bandwidth-intensive games at over 50 MB per hour of play. However, FPS games are played by a minority of gamers. The most popular games, such as casual online games, non-shooter console games, and multiplayer virtual worlds, are surprisingly lightweight in terms of traffic, partly because much of the graphical processing and rendering occurs on the client side, which is downloaded once and then does not cross the network again after download. There is a possibility that multiplayer games will break into the mainstream, whereas we assume that online gamers constitute a maximum of approximately 40 percent of the online population, and that multiplayer gamers are a maximum of approximately 40 percent of the gaming population. Please see the companion paper "Approaching the Zettabyte Era" for consideration of this and other alternative scenarios.
