There's a new issue brewing in the upstream path, and
it has nothing to do with signal ingress or impulse noise.
Instead, it's traffic management -- how to manage all
of the services that are supposed to ride in the 5- to 40-megahertz upstream-signaling
path.
The questions of how and when to become broadband-traffic
specialists, MSO and vendor technologists said, must at least be posed this year.
That's because more and more advanced services are
poised to crowd the already-slim upstream path. True, penetration rates for high-speed
data and fledgling telephony services aren't large enough on an industrywide basis to
command immediate attention. But they will be, if MSO intentions become reality.
"As you do high-speed data, and then stack on
telephony -- whether switched or IP [Internet protocol] -- at that point, traffic patterns
will have to be watched very closely," noted Paul Gemme, vice president of plant
engineering for Time Warner Cable.
"This is the year to understand traffic modeling and
engineering, as penetration rates increase for all of the advanced services," agreed
Susan Marshall, senior vice president of advanced products for TCI.NET, the
high-speed-data arm of Tele-Communications Inc.
Marshall said TCI is already working with @Home Network to
tackle traffic issues, but the work is fairly preliminary.
Traditionally, cable-broadband operators haven't
needed to be overly concerned about traffic patterns, because those issues were handled
architecturally. Bandwidth upgrades handled downstream-channel expansions, and few
services shared the 5- to 40-MHz upstream path.
That's in sharp contrast to the telephone industry,
which has long used traffic engineering to establish how much switch capacity must be in
place to adequately serve the communications needs in neighborhoods linked to central
offices around the world.
"There will be another year or two of bandwidth
abundance in the upstream, which pretty much makes 1999 the year that the industry should
start asking, 'Then what?'" said Andy Paff, chief technology officer for
WorldBridge Broadband Services Inc., a Lakewood, Colo.-based firm that specializes in
upstream-path communications.
But fast-forward to a home of the not-too-distant future,
and the resounding need for broadband-traffic management becomes more readily apparent.
Consider this scenario: Digital-video services, both
broadcast and on-demand, are flowing over multiple digital set-tops to as many TVs. Four
different family members use cable modems to access the Internet. The family has also
subscribed to a primary and secondary telephony package from their cable operator, which
gives them four lines of service. That's good for the two teen-agers in the house,
who can chat with their friends for hours while they surf the Internet, without
interrupting calls to their parents.
They've also subscribed to a video-telephony option
for their telephony package, so that they can call and see geographically separated
friends and relatives.
From a revenue perspective, it's a cable
operator's dream. Using estimated prices, that home is easily forking over $150 per
month to its cable-broadband provider.
But if that home falls into the "mainstream," and
not the "early adopter" category -- meaning that 99 or so other homes served by
the same 500-home node are taking identical services -- it quickly goes beyond a good
problem to have, particularly without traffic-management techniques in place.
"You don't want to wake up one day and find out
that, uh-oh, there's no bandwidth for those two customers that you're supposed
to install," Gemme noted.
Even though nodes can be split in less than half a day to
free up additional upstream spectrum, "it's still a situation that we'd
obviously rather preplan, rather than having a panic point," Gemme added.
Identifying the bandwidth hogs is part science and part
black art, engineers said. The easy-to-spot gobblers are video telephony and some
work-at-home data applications. With video telephony -- which is not a broadband reality
today -- a camera is attached to the set-top or personal computer and a bandwidth-chomping
video session is established from the home.
Telecommuting users become problematic, from an available
bandwidth perspective, when users send large files upstream to coworkers over the
broadband plant, instead of sending small requests upstream and receiving the bulk of
transmissions in the downstream direction.
What clouds matters is a complicated mix of prediction,
network architecture, service-penetration rates, simultaneous usage and individual service
characteristics.
What happens if upstream channels get clogged up with too
much stuff? Phone calls could sound not unlike cellular phones in bad coverage areas, and
high-speed-data transmissions could experience delay-related glitches.
There are currently two ways to deal with the need for
additional bandwidth in the upstream path, Paff said: brute force and strategy.
"Brute force is shrinking the size of the node;
strategy is routers and enhanced intelligence in the IP network," Paff said, adding
that the latter "will take at least a year to develop."
Executives with TCI and Time Warner agreed that there is
little on the drawing board these days to address the looming upstream-bandwidth crunch.
But both MSOs said they could foresee solutions that tapped into cable-modem-headend
systems, where most IP intelligence is currently anchored.
"At this point, it's on the radar screen"
only, Gemme said.
Paff said the cable-broadband industry would soon be able
to cash in on developments made by the alternative long-haul carrier segment.
"Companies like Level 3 [Communications Inc.] and
Qwest [Communications International Inc.] have a more urgent need for packet
management," Paff noted. "Finding a way to intelligently route packets to manage
bandwidth is a core issue facing that industry, and a lot of usable technology will come
out of their needs."
