Can ATM and IP peacefully coexist in tomorrow's integrated voice and data market? Actually, they
already do
internettelephony.com
Voice over IP; Voice over ATM; Voice-over stars
GREG LANGDON
The public network does an outstanding job serving the public's needs for voice communications.
The system is distributed widely, well-understood and highly reliable, and it delivers voice with
exceptional clarity. But as with any legacy system, there is room for improvement. The desire to
reduce costs, use infrastructure more efficiently and integrate voice and high-speed data
transmission have fueled intensive efforts to develop alternatives to the public network.
These efforts focus on transmitting voice over networks that initially were deployed to transmit data
and fall roughly into two camps: those focused on IP networks and those focused on ATM networks.
These camps often are seen as competing technologies, giving rise to heated debates as to which
is the superior method for voice transmission over a data network: voice over IP or voice over ATM.
In truth, voice over IP and voice over ATM each are suited to different applications and operating
environments and are not head-to-head competitors. Because of inherent characteristics, including
the underlying technical aspects of IP and ATM networks, voice over IP and voice over ATM likely
will coexist peacefully for some time as an alternative means for voice transmission over the public
network.
Voice over IP: Apps high and low
Because of intense interest in new Internet applications, voice over IP is more widely discussed
than voice over ATM. The attention is understandable. Given the runaway popularity of IP-based
networks, it makes sense to explore using these same networks to transmit voice.
A best-effort delivery system originally developed for Internet addressing and routing, IP uses
variable packet lengths of 1 kbyte or more. Because of the size of these packets, they take time to
fill, and when a packet is lost, substantial time is required to resend it. The use of large packets
benefits data transmission by increasing payload relative to protocol overhead, and the associated
packet delivery delays cause no real concern in data transmission.
This is not the case with voice transmission, however, because even small delays can have a
negative affect on voice quality. Unlike voice over the public network, which typically proceeds
much like a face-to-face conversation, voice over IP based on current standards suffers from noise,
delay, echo, jitter and packet loss.
These conditions can resemble those that formerly existed on overseas calls via satellite links, when
delays caused confusion between participants as to whether the other party was continuing to
speak. As a result, voice over IP largely has remained focused on those willing to endure
transmission delays and poor audio quality as a trade-off to gaining a low-cost alternative to toll
charges.
This situation exists because of quality of service (QOS). H.323, the standard protocol currently
used to control voice over IP, defines telephony gateways, gatekeepers and the procedures for call
initiation and termination, but it cannot guarantee QOS. Efforts at improving H.323's ability to
support voice over IP on the Internet, including the development of protocols such as MGCP and
SIP, are designed to address issues of extensibility and scalability, but even then QOS issues
remain. Further definition and development of standards will be required before the industry can
envision anything approaching an IP-based public telephone network as shown in Figure 1.
To hasten the arrival of near toll-quality voice over IP, several companies have developed
proprietary protocols as alternatives or extensions to H.323 with a specific focus on QOS. Because
of their proprietary nature, these approaches lack interoperability with other designs and require
that both ends of the connection use the same type of hardware and software. Such solutions are
useful in private links between corporate branch office systems--for example, to interconnect PBX
systems between separate offices. Although they represent promising solutions in dedicated private
links, such proprietary solutions do not move us toward a broad-based application of voice over IP.
Voice over ATM: High-quality apps
Voice over ATM actually predates voice over IP, although relatively few people were aware of it
before the advent of DSL. ATM's emergence as the standard Layer 2 transport protocol for DSL has
brought ATM to small and medium-sized businesses and homes, and the recent introduction of
integrated voice and data over DSL marks the first broad application of voice over ATM.
Voice over ATM provides high-quality transmission of voice traffic by taking advantage of ATM's
inherently voice-friendly nature. ATM is designed to carry multiple and varied traffic types
simultaneously and provides exceptional traffic management and the ability to deliver service
classes based on traffic type.
Important for voice transmission, ATM also provides guaranteed QOS.
Unlike IP's variable packet sizes, ATM uses small, fixed-length data packets of 53 bytes each. Small
packets fill and transport more quickly and are therefore less susceptible to network delays. The
smaller packet size used by ATM typically results in voice with much lower delay than that of an IP
network and a more natural quality for conversation.
However, ATM does not reach the desktop as IP does, so voice over ATM probably will not bring
services directly to a handset or computer the way voice over IP may some day. Instead, voice over
ATM likely will serve businesses and networked homes as it does in the DSL integrated voice and
data services being tested today. This type of service uses voice over ATM to deliver voice over
DSL, integrated with data, to a next generation integrated access device that handles all a facility's
voice and data traffic (Figure 2).
To date, DSL service has been aimed at customers who want high-speed Internet access. This type
of service adds high-speed data functionality to an existing copper pair without degrading the
voice service also carried on that line, but it does not expand telephone services. Voice traffic still
is carried as usual over the public network, and adding to voice capacity still requires adding more
copper lines--at considerable expense.
Newer DSL services that integrate voice and data provide the same high-speed data functionality
while using voice over ATM to expand voice capacity. This is a significant extension of services
because voice over ATM can deliver up to 24 toll-quality voice lines over a single twisted pair.
Underlying technical aspects of ATM, specifically the ATM adaptation layer (AAL), provide this
functionality.
The AAL defines how voice and data traffic is converted into ATM cells and translates such higher
layer services as TCP/IP into the size and format of the ATM protocol layer. Several adaptation
layer definitions exist, and each accommodates a network traffic type. The two AALs most often
used for voice traffic are AAL1 and AAL2.
Voice over ATM with AAL1 is the traditional approach for constant bit-rate (CBR), time-sensitive
traffic such as voice and video, and it provides circuit emulation for trunking applications. While
ATM with AAL1 is suitable for voice, its use of fixed bandwidth allocation means network resources
are consumed--even when no voice traffic is present--until the connection is terminated.
A more recent definition, AAL2, uses statistically multiplexed variable bit-rate (VBR) ATM traffic
classes to provide significant improvements over AAL1. AAL2 supports both CBR and VBR
applications and delivers multiple voice calls over a single ATM permanent virtual circuit. AAL2
traffic also carries content information used to prioritize network traffic and provide for dynamic
bandwidth allocation.
The dynamic bandwidth allocation of AAL2 provides a significant advantage to carriers and
adopters of integrated voice and data over DSL. Dynamic bandwidth allocation results in full use of
available bandwidth, with ongoing and transparent reallocation as demanded by network traffic.
This enables increased data throughput when voice usage is low or during periods of silence in a
conversation, with data throughput automatically throttled back when bandwidth is needed for
additional voice traffic.
A tag-team effort
The quest for economic and technical benefits continues to drive the exploration of data networks
as alternatives to the public network for voice communications. The challenges of carrying voice
traffic on data networks include scalability, compatibility with existing protocols and equipment
and, most notably, QOS. As new developments address these issues for voice over IP and voice
over ATM, each is finding its place in applications for which underlying technical aspects make it
most suited.
Today, voice over IP has value in low-cost Internet applications that do not require high QOS
levels or connecting high-cost, private links between corporate branch office systems. Voice over IP
also has appeal from an applications standpoint; it consolidates services as it allows telephones
and PC applications to run on the same network. It eventually may lead to unified messaging for
voice, fax, e-mail and other services.
Voice over ATM provides value primarily in the delivery of integrated voice and data services over
DSL. With its role in enabling multiple, toll-quality voice lines plus high-speed data over a single
copper pair, voice over ATM helps move DSL beyond its already prized capability for providing
high-speed data transmission.
Voice over IP and voice over ATM already coexist, and their use is beginning to enter the
mainstream. As they grow in popularity, voice over IP and voice over ATM will continue to coexist,
with each serving the applications to which it is inherently best suited.
Greg Langdon is Vice President of Marketing for Efficient Networks, in Dallas. His e-mail address is
glangdon@efficient.com . |
