< I'd be interested in knowing what methodologies were used to measure this traffic>
Internet Traffic Soars, But Revenues Glide: Definitions and Methodology - May 6, 2002
RHK’s report Internet Traffic Soars, But Revenues Glide, presents our key findings on Internet traffic and revenue growth in the United States. We examine the drivers behind the continuing expansion, with projected growth to 2004, as well as the factors constraining associated revenues.
Internet growth adds extraordinary volumes of traffic onto backbone networks each year – monthly traffic now stands at around 100 Petabytes per month, more than twice the equivalent long distance voice traffic for the entire nation. An in-depth understanding of this growth, its sources and likely development, is essential for service providers that are planning the next-generation networks of the future.
The following sets out the definitions and methodologies RHK uses to quantify Internet traffic and to interpret the underlying patterns and trends.
Definitions
RHK defines Internet traffic as that traversing the dedicated Internet infrastructure, rather than IP-denominated traffic that passes through intra-company leased lines or the public network. ATM and Frame Relay traffic are also excluded in this definition.
Internet traffic is further defined as the total of loads offered to carriers in the form of entry traffic to the network originating from end users and peering partners. This is distinct from the traffic carried by a backbone provider, which is the sum of the traffic on each link. Therefore, the offered load equals carried load divided by the average number of hops in a typical path.
Figure 1: Carried load vs offered load
Offered load = åL
Carried load = åC
Offered load = Carried load/Average # hops
Capacity is defined as the available bandwidth, that is, the total of available access lines reported in bits per second. For 2001 and 2002, RHK estimates the ratio of capacity to average traffic as 10:1.
Peak traffic is reported in the busy hour in bits per second. Peak traffic is projected on a monthly basis, assuming the same traffic profile for every hour of the month at the peak rate. For instance, if the peak hour is 9.00am, the average busy hour is the average for 9.00am over the month. For 2001 and 2002, RHK estimates the ratio of peak to average traffic as 1:6.
Average monthly traffic reported is based on an average hour projected over a month.
Only average traffic is considered in this report as we believe this provides the best understanding of the demands placed on Internet infrastructure. Metrics such as peak traffic, latency, jitter and packet loss are useful in addressing issues of network design, quality of service and SLA-compliance, however, they are less indicative of ongoing traffic patterns and so are not considered in this report.
RHK defines average traffic as the amount of data that comes in, or goes out of, a router port over a certain period of time. For example, the following formula converts from samples taken at five-minute intervals in Gigabits per second to average traffic expressed in Terabytes per month.
T = [(å samples over a month)/ (12 x 24 x 30)] x [(3600 x 24 x 30)/ (8 x1000)]
Where:
Ave. sample in Gigabits per second (a) = (å samples over a month)/(12 x 24 x 30)
Average sample in Terabytes per second (b) = a/(1000 x 8)
Average traffic in Terabytes per month = b x (3600 x 24 x 30)
Therefore:
T = (å samples over a month) x 3/80 Terabytes per month.
Methodology
Sources
Public and private data were provided under non-disclosure by network engineers at the top backbone providers. In total, these sources represent companies handling more than 60% of the total Internet traffic of Internet traffic in the United States.
TNS Telecoms interviewed 3,500 business locations on RHK's behalf to determine the percentage bandwidths used for voice, private line/ATM/frame relay, and Internet.
Federal Communications Commission’s ARMIS database of access bandwidth growth.
RHK’s broadband subscriber forecast, Broadband Access: Global Market Forecast, April 2002 provided projections of business and consumer take-up of xDSL and cable modem services.
Measurements
Traffic growth is reported, analyzed and compared from three perspectives, as experienced by: backbone carriers; data centers; and residential and business users.
The traffic figure was calculated by totaling the number of ports sold by the top backbone service providers and the average traffic on those ports.
Revenues were totaled for wholesale dial-up, business dial-up, dedicated access and hosting. Average revenue growth was weighted by service provider market share (by revenue).
Assumptions and data points
Before 1999, most of the world peered with US carriers, so close to 100% of worldwide traffic passed through US backbones. Today there are more peering points outside North America. RHK estimates 66% of worldwide traffic passed through the US in 4Q00, falling to 60% in 4Q01 and 55% in 4Q02.
In 2001, the top seven backbone service providers captured at least 55% of total US Internet traffic.
Residential broadband subscribers (xDSL plus cable modem) in North America will reach 36.8 million by 2005.
Dial-up is the access method of choice for 79% of residential subscribers. However, bandwidth usage is dramatically different. Cable subscribers account for 14% of total subscribers (cable, xDLS and dial-up) but as much as 73% of bandwidth, mostly due to users running web-servers off their connection.
Best to visit this URL for graphs/diag:
networkwatch.rhk.com
Stang |
