IETF defines network capacity
- By Joab Jackson
- Feb 29, 2008
Here's a quick riddle: What is the bandwidth of a fully functional 802.11b wireless link? A network administrator might say 11 megabits/sec, while the end-user would observe 8 megabits/sec, and an electrical engineer would say 25 Mhz.
And they would all be correct, at least according to their own definitions of the term "bandwidth." But when two parties get together to talk about network capacity, such as a service provider and government agency, how can they ensure they are talking the same numbers?
Now two members of the Internet Engineering Task Force have proposed formal definitions of network capacity, a task that until now has, remarkably enough, gone unaddressed. The authors hope these definitions will become industry standards for defining bandwidth.
Philip Chimento, a senior professional staff member at The Johns Hopkins University Applied Physics Laboratory, and Joseph Ishac, a computer engineer at the NASA Glenn Research Center, authored the Request for Comments 5136
, which addresses the issue.
"There are really a lot of tools that probe the network and measure something called 'bandwidth'. But they don't all return the exact same information, and the results could be interpreted differently," Chimento said,
The work is important insofar as different measurements of bandwidth could lead to disputes among different parties, such as network providers and their customers, over the actual capacity under discussion. "Having some common definition allows everyone to talk the same way about the same quantity," Chimento said.
Measuring bandwidth may be trickier than it may first appear. "Measuring capacity is a task that sounds simple, but in reality can be quite complex," the RFC states.
For instance, radio-frequency wireless networks use error-correcting code and retransmissions at Layer 2, the data link layer of the Open Systems Interconnection 7 Layer Model of networks. This is done to ensure packets arrive at their destinations. While these procedures may use some of the available physical capacity, they are not factored into bandwidth estimates at the layer above the data link layer, or Layer 3, the Internet Protocol (IP) layer of the OSI model.
"What we're focusing on is what capacity is available to the IP layer, not the total physical capacity, which both Layer 1 [the physical link layer] and Layer 2 might use some of for control and housekeeping and those kinds of functions," Chimento said.
The users of these definitions can determine for themselves if they wish to include the control packets at Layer 3.
The engineers preferred to jettison the term "bandwidth" altogether which, the RFC states, is overloaded with conflicting meanings. Instead, they have come up with working definitions for two terms, "capacity" and "available capacity," in relation to traffic between two points in a IP network.
Moreover the authors have provided a way to couple these terms to measurements of time'namely, allowing users to specify when the measurement was taken, and the time interval of the measurement taken.
"Those two pieces of information add a lot of meaning to measurements you make of a network," Chimento said. "If you measure network capacity at a particular point in time, it's hard to make valid inferences from that single point to what you're seeing over a period of time."
When taken together, the time interval and point in time when the measurement was taken will allow users to make more valid inferences about the actual capacity of the network being studied, Chimento said.
Although labeled a Request for Comments, the document is in its finalized form, Ishac said. The authors hope that IT vendors and service providers will adopt these definitions in their tools and discussions.
"Then people can have some common ground and people can understand one another when talking about talking about capacity," Chimento said.
Joab Jackson is the senior technology editor for Government Computer News.