The force behind fragmentation

For many providers, especially those that concentrate on offering DSL, tacking voice onto their services suite is one way to keep customers. But voice over DSL still has problems such as jitter and latency. Coupling those problems with issues such as uncertain scalability and reliability has forced many providers to refrain from rolling out voice over DSL.

To counteract problems associated with the technology, the FRF.12 standard was developed a few months ago. Adaptive fragmentation segments traffic only when voice is present and only when it is needed to compensate for voice delay tolerances. Both ATM and non-adaptive fragmentation, which renders poor voice quality, cut up data frames into short, fixed-length cells to ensure low-delay, toll-quality voice.

"In order to do voice over DSL, you must do fragmentation, and adaptive fragmentation is the best way to accomplish that," said Stefan Knight, director of product marketing for CopperCom. "It is essential."

Now technologies are appearing in support of the FRF.12 standard. "This came about because of service providers' interest in deploying derived voice services over DSL to customers," said John Reister, assistant vice president of advanced technology at Copper Mountain. "But the challenge is really at the lower line rates."

As the line rate decreases, delay and jitter problems arise. A long data frame will travel on a 1.5 Mb/s DSL connection faster than it will on a 160 kb/s DSL link. In theory, little fragmentation is needed at the higher speeds, compared with the lower speeds, where large data blocks should be parceled into many fragments.

"We take a look at the line rate, and if it's at a megabit, you don't need fine chopping, you might just chop it in half," Reister said. "You always want to fragment only what you need to and not any more, so you minimize jitter without added overhead" (see figure).

But some say adaptive fragmentation is needed at those line rates. "Most customers aren't trying to deliver voice at the lower rates at this point," said Byron Young, senior director of product management at AccessLan Communications, which has a similar technology based on the FRF.12 standard for its equipment.

Still, providers such as Jato Communications are interested in the benefits of adaptive fragmentation. "As we add services to our network, we want to differentiate on [quality of service]. Adaptive fragmentation is a way for us to do that," said Mitchell Ashley, vice president of network integration at Jato. Jato currently uses Copper Mountain equipment with the adaptive fragmentation software.

Onsite Access, another provider that uses Copper Mountain's equipment to deliver services to the multi-tenant unit market, also plans to use the adaptive fragmentation technology. "This takes big steps in solving problems we have had with the timing of voice traffic," said Brandon Knicely, chief technical officer of Onsite Access. Onsite Access currently has been running voice traffic on a time division multiplexing network tocomplement its DSL service.Jato's Ashley hopes more equipment vendors adopt ad aptive fragmentation technologies. In addition, having similar technology in other network elements such as IP routers would be valuable, he said.

But AccessLan equipment currently can support "adaptive" or "dynamic" fragmentation, Young said. "This is a capability that absolutely should be part of the equipment."