VoIP Scalability

Telephony was the dog that didn’t bark (too loudly) at Cable-Tec Expo this year. In other words, there was significance in its relative silence.

The significance is that the industry at large has successfully deployed voice over IP technology, at least on the residential front. Business voice is another matter, as is telephony for the largely forgotten small, one-way systems.

The very success that the major cable operators have had with VoIP, however, leads to the important topic of scalability, which was the subject of three separate technical papers and presentations offered at Expo. Here’s a summary. Managing growth MSOs added approximately 2.5 million VoIP subscribers in 2006, bringing the total number to 8.6 million subscribers. Citing these statistics, Dan Rice, vice president OSS product development at C-COR, said in a presentation that this represents a 51 percent increase compared to 2005.

"There is every reason to anticipate ongoing increases in the subscriber count at levels on par with the (number) gained in 2006," he wrote in his Expo paper.

Hence, MSOs will need to scale capacity in order to avoid congestion, which in a DOCSIS situation could affect not only voice calls, but also data services. "The blocking probability for DOCSIS is a calculation of the probability that best-effort traffic will be unacceptably impacted by adding QoS-enabled VoIP," Rice wrote.

Rice suggested adapting the traditional Erlang method to aid with VoIP capacity planning. He offered several steps, including calculating available capacity and the bandwidth for a single unsolicited grant service (UGS) flow for a single call, using the number of simultaneous flows that can be supported as the "n" in Erlang and comparing the solution to the number of MTAs. Scaling VoIP Another presenter, Bil Dry, a technical leader for Cisco Systems, reminded attendees that when capacity planning, an MSO must consider every VoIP device and process. "Can the CMS (call management server) accommodate the next rate center addition?" he asked as an example.

He offered several points to keep in mind when scaling a VoIP network with an eye on cost. With a little facility planning, for example, the gear that produces the most heat can be located in the coldest portion of the point of presence (POP) to avoid extra cooling expenses.

Calculate downstream utilization when adding upstreams, Dry reminded. "The downstream DOCSIS channel may be a scaling bottleneck if the MAC domain exceeds four upstreams per downstream." Remember that changing from quadrature phase shift keying (QPSK) to 16-QAM (quadrature amplitude modulation) almost doubles simultaneous VoIP calls per upstream, he added. "Consider enhancing plant quality to support DOCSIS 2.0 64-QAM upstreams." High growth Geography is an important dynamic when scaling, said Parviz Rashidi, VoIP and IMS network architect, network business solutions, Nortel, in his presentation. "Some regions may have dense populations, while others may be thin in population or may be hundreds of miles away from other parts of the network."

He discussed several models, including growth by re-homing and overlap. For the first, there is a separate CMS for each of two sub-networks, divided based on subscriber distribution. In the second, control overlaps between two or more CMS/MGC (media gateway controller) nodes. Growth is accommodated by one CMS to a certain point and then switched to the alternative.

Rashidi also addressed routing scalability solutions, including tandem switching, SIP redirect/routing proxy and database lookup, and using ENUM as a database.

"Either the CMS switch can do the DNS ENUM query directly to get the routing information, or the ENUM server can be the SIP proxy," he said, noting that the overlap method thereby becomes "a favorable option" in "high growth" areas or where "regional/administrative differences" do not matter. – Monta Hernon and Jonathan Tombes