Cable engineers and vendors are working with Cable
Television Laboratories Inc. to stretch the performance capabilities of first-generation
MCNS modems to new levels that include support for packet telephony and office services.
The question still to be resolved, they added, is whether
software-only approaches will meet the service goals being set by operators. If not,
adjustments in the hardware could affect the designs of some, though not all, suppliers of
first-generation MCNS modems.
The initiative -- spearheaded by what is known as the
"QOS (quality of service) Extension" group within the Multimedia Cable Network
System process -- is closing in on software solutions that would enable QOS capabilities.
Cable engineers hope to implement those changes by the time MCNS modems are deployed
commercially early next year.
"From the 50,000-foot level, I'm pretty confident
that we can accomplish what we're trying to do within the framework of DOCSIS [Data
Over Cable Service/Interoperability Specification]," said David Fellows, chief
technical officer of MediaOne Express and chairman of the Society of Cable
Telecommunications Engineers' subcommittee on modem standards.
"Any additional hardware hooks that we need will make
it into the first generation," he added.
So far, the QOS Extension task force has been focused
strictly on software improvements, and it is finding ample reason for the optimism
expressed by Fellows, said Rich Woundy, chairman of the group and chief architect at
American Internet Corp., a Bedford, Mass.-based supplier of data-provisioning software.
Woundy's group has been working on bringing together
the best elements of several proposals for adding QOS capabilities to MCNS modems.
He said companies such as Motorola Inc., Bay Networks Inc.,
3Com Corp., Turbonet Inc., Cisco Systems Inc., Com21 Inc. and General Instrument Corp. are
making key contributions to "what is now becoming a consensus specification."
"Our next step is to prepare a white paper that will
serve as an engineering-change proposal to be submitted to the DOCSIS group," Woundy
said.
The QOS effort is aimed at meeting goals set by the
industry's Packet Cable initiative, which is spearheaded by a CableLabs task force
headed by Mark Coblitz, vice president of strategic planning at Comcast Corp. The group is
trying to set service parameters for IP (Internet protocol) telephony and other
capabilities, such as videoconferencing, video-streaming and multicasting.
Woundy said a major goal is to ensure that IP telephony
works in conjunction with MCNS modems without disrupting MCNS implementation.
"We think that we can solve at least 80 percent of the
issues related to these QOS goals within the current parameters set by DOCSIS," he
said.
As described by Fellows, there are three areas of activity
where "hooks" added to first-generation MCNS modems -- mostly by adding software
at the headend -- can accomplish the QOS goals.
Advanced protocols that set bit-rate priorities and
continuous packet flows can easily be implemented over routers to accomplish QOS in the
downstream to the MCNS modem. But setting such parameters over the MCNS return channel --
which is designed to deliver packets on a contention "best-effort" basis -- is
much more difficult.
One key area of work involves a means of allowing the CMTS
(cable-modem termination system) at the headend to authorize a particular modem to send a
constant bit stream in the return channel, such as what would be required in a telephone
call, versus operating in the default mode on a contention basis. This would be done on
the basis of the CMTS determining what the modems in a particular serving area have in
queue in the way of data waiting to be sent over the upstream channel, and then assigning
a mini-frequency slot for the constant-bit-rate transmission requested by the voice caller
or callers.
Along with this capability, the group wants to be able to
dynamically assign QOS authorization to particular users as they request service, which
would allow cable operators to bill for specific usage, rather than establishing locked-in
monthly rates for various service tiers.
"Dynamic assignment of service IDs would allow us to
assign a particular user with a higher priority for the time frame needed, after which the
user might revert back to best-effort status," Fellows explained.
These are the pieces that Woundy's group is focused
on. The third piece, labeled "fragmentation," involves the ability to break up
long data packets to create more space for higher-priority data streams that might involve
much shorter packets, as in voice communications.
This might entail breaking the long packets into cells
commensurate with the space available within a given mini-slot, Fellows said.
For modem manufacturers that are implementing MCNS using
ASICs (application-specific integrated circuits), adding the fragmentation capability may
not require any refinements in hardware, Woundy said, noting that the fragmentation
effort, to be taken up by another group, is "very much hardware-specific."
However, suppliers depending on CPUs (central processing
units) in conjunction with multiple software systems to implement the MCNS capabilities
"could run into trouble" when it comes to achieving fragmentation, he said.
It will be difficult to provide the QOS level of support
necessary for toll-quality telephony using the "existing generation of DOCSIS
silicon," said Buck Gee, vice president of marketing at Com21, which is marketing
modems that employ cell-based ATM (asynchronous transfer mode) technology.
But, he added, it might be possible through adjusting the
latency requirements a bit to offer an acceptable range of services within the QOS
parameters that are doable in software.
"There will be a market for second-line, fax and other
services that aren't quite toll-quality," Gee said.
Bob Schack, director of marketing at Cisco, disagreed,
contending that software has advanced to the point where appropriate implementation of
protocols throughout the network, as well as through the CMTS to the modem, will ensure
toll-quality calling and many other capabilities over MCNS systems.
"We're completely neutral on cell-based versus
packet technology, because we have both types of product lines and, if people want to go
to cell-based systems, we will be happy to support them," Schack said. "But we
don't think that's necessary."
The big question is what type of voice and other services
does the industry want to provide using the QOS capabilities that are now available,
Schack said.
He noted that developing the right feature sets is a top
priority with Cisco and its cable customers, including MediaOne Express, which announced
last week that it signed a contract to develop advanced end-to-end broadband
infrastructures capable of delivering digital voice and video, as well as data services.
But if Woundy is right, the feature set that operators have
to work with in conjunction with at least some first-generation MCNS modems might fall a
bit short of the full field-of-dreams aspirations of the cable industry. The good news is
that with such niches as video telephony, high-speed-data streaming and second-line voice
service wide open to cable, the Packet Cable infrastructure could prove to be a major
revenue-generator, even at the 80 percent level of QOS implementation.
