|
CABLE WORLD STAFF
With penetration rates for high-speed cable modem service soaring into double-digits and more packet-based services on the near horizon, the need to predict accurately the flow of data packets across cable networks is ever more crucial.
To meet this need, CableLabs[R] is developing an analytical tool for its MSO members - so any current and future traffic patterns can be identified and planned for at the cable system level.
Economically, a predictive data modeling tool is a big plus for cable system finances. Because of backbone costs, it's undesirable to buy bandwidth before it's needed. Because of long (and lengthening) lead times for backbone capacity increases, waiting too long also can be unattractive.
"Continued growth of high-speed data services will require capital expenditures to provide satisfactory performance levels, which is why a simulation tool is helpful," says Dr. Terry Shaw, senior adviser-network systems at CableLabs. "It simply helps to manage growth."
Cable's major operators added about 65,000 high-speed data customers per week in the second quarter of this year, ending the period with roughly 2.2 million subscribers. Growth is anticipated to continue at an aggressive rate. Meanwhile, high-speed data customers are increasingly savvy with high-bandwidth applications, such as streaming media, which eats bandwidth voraciously.
The tool, originally created for internal CableLabs testing of cable modem equipment, works by simulating a wide range of common plant and configuration parameters, such as node size, fiber distance, last mile and equipment type. Once a "typical" system is virtually constructed, the modeling begins. More and different types of data traffic get piled on to determine just what makes a cable data network clog - and when. The tool is designed to work on PC and Sun computers.
"It's a non-invasive way to determine how to best configure a high-speed data network," explains Shaw. "Without something like this, you're potentially changing things on an active network, which could affect customers."
The model handles data loads from homes to headend gear and outwards to high-speed backbones, so traffic capacity can be predicted all along the chain. It also helps determine what "size" backbone is needed for varying penetration rates, which is fiscally sensible.
The tool simulates different types of data users, from simple e-mail and Web surfers to networked homes pulling down huge files or participating in a streaming event. The traffic types can be examined on a per-node or network-wide basis, aggregating all nodes for a complete hot-spot analysis.
In one scenario, Shaw selected eight seconds as the "threshold of pain" at which data customers would stop waiting for a Web page to load and was able to examine what network conditions would exceed that limit.
"We're very rapidly evolving from a single-user, single-service platform to a multiple-user, multiple-service platform," Shaw explains, referencing home networking advances. "Under the simulation, the DOCSIS platform performed very well, but we wanted to stress it to see what would happen when more than one user shares a single cable modem."
The tool is under development and isn't ready for use outside of Cable-Labs' Louisville, Colo., headquarters. The project is high up on the to-do list, though, particularly because of strong cable interest in putting it to use.
Back to this issue
|
