from Telephony magazine, Oct 98:
The rebirth of ATM
ALEX DOBRUSHIN
The rumors of ATM's demise have been greatly exaggerated.
While taking a bashing in the local area network, the use of asynchronous transfer mode (ATM) in carriers' wide area networks has been growing
steadily--and is expected to reach $5 billion in equipment sales by year 2001, according to Dataquest, a San Jose, Calif.-based market research firm.
Even IP-centric carriers, like Level 3 Communications, Inc. (Omaha, Neb.) and Qwest Communications Corp. (Denver) committed their backbones
to ATM, driven by its ability to provide scalable infrastructure that can efficiently transport all offered services. While IP is being positioned as the
end-to-end networking service protocol, ATM has been chosen for the next generation of transmission and multiplexing.
A large part of the Internet depends on ATM for its operation, and virtually all networks delivering frame relay, the most popular data service, use
ATM as its backbone. With frame relay and IP services also reaching billions of dollars in revenues, ATM once again is coming into the spotlight.
As part of ATM's success, carriers are looking to integrate ATM with their transmission infrastructure. An example of this movement are the
announcements by Ascend Communications and other ATM vendors that tout integration of Sonet transmission functionality into their switches,
allowing them to be connected directly on fiber.
By bypassing traditional transmission gear such as large Sonet add/drop multiplexers (ADM) used in a network backbone, carriers can achieve large
operating savings while simplifying provisioning and administration of services. With backbone networks in active deployment, carriers are now
extending ATM benefits to their access transmission networks with ATM-based Sonet add/drop multiplexers.
It's all about services…
In the local access networks, where traditional time division multiplexed Sonet is a dominant method for transporting bandwidth to business subscribers
today, ATM can solve a critical issue for service providers. The issue is simple: Carriers are missing out on highly profitable packet service
opportunities because their Sonet transmission infrastructure was built around fixed bandwidth services such as circuit switched voice.
Traditional Sonet only delivers bandwidth in fixed increments, like DS-1 (1.5 Mb/s), DS-3 (45 Mb/s), STS-1 (155 Mb/s) and other similar slices
defined by its rigid synchronous transfer mode (STM) multiplexing hierarchy. The data services do not map well into these fixed bandwidth slices. For
example, even a simple packet service, like transparent LAN service (TLS), requires myriad of network elements and signal conversions.
Typically, a carrier-owned router converts customer's 10 Mb/s Ethernet connection to a 45 Mb/s ATM DS-3 interface that is then carried over the
Sonet ring's 52 Mb/s STS-1 circuit (Figure 1). The result is that 42 Mb/s (80%) of available transmission capacity is wasted. What's worse is that
each subscriber will need a discrete hand-off port in the carrier's CO, occupying expensive switch ports and wasting even more expensive switching
capacity.
This approach to provisioning a service is by no means economical or simple. Such kluged solutions also take a lot of planning and are slowing service
rollout. In today's highly competitive environment, service differentiation with this approach is nearly impossible.
The long-term success of any carrier depends on its ability to attract business customers looking for added value from carrier services. Business
subscribers' needs are getting more sophisticated each day.
In the past, they sent simple, text-based e-mail. Today, they are adding multimedia attachments, streaming audio and video, and are actively engaging in
electronic commerce. They also are starting to experiment with other applications such as packetized voice over IP. For any carrier to win, a new "data
aware" Sonet infrastructure is essential to enable service creation required to profitably fulfill these needs in the next century.
ATM to the rescue
ATM was invented to provide a simple, yet robust, multiplexing protocol that could transport all services with high efficiency. It is no surprise that it is
finding its way into Sonet access networks.
With ATM, it is possible to create full service access networks that can effectively support traditional fixed bandwidth services while enabling virtual
private network (VPN), Internet Protocol (IP), asynchronous transfer mode (ATM) and other variable bandwidth services to be easily delivered to
business subscribers over the carriers' existing optical infrastructure. The concept is simple: Replace Sonet's antiquated STM multiplexing with ATM
and you have the enabling infrastructure for all services.
One of the critical requirements for Full Service Sonet is that it must also perform on par with traditional transmission networks to achieve credible
support for existing fixed bandwidth services without affecting quality of service (QoS) or service survivability. This is where blending Sonet and an
ATM capability creates a perfect match. The key, of course, is the word, "blending." For Full Service Sonet to be rapidly deployed, it must meet
certain carrier operating objectives.
Because carriers have invested billions of dollars in their Sonet infrastructure, they need to know that this new solution interworks with other pieces of
their optical network, including higher layers of traditional Sonet and dense wave division multiplexing (DWDM).
Such key Sonet ring aspects as the low pass-through latency and sub-50 millisecond healing also must be preserved so that service performance,
engineering and operations of embedded voice switching infrastructure are not affected.
The first aspect is easily addressed. There is a clear standards definition about how to map ATM cells over the Sonet physical layer. The latency and
healing aspects in the ATM ring are also being addressed by standards organizations.
These parameters aside, ATM has a clear advantage over other solutions because the standards for services mapping are well defined. Let's look at a
few examples.
Traditional T-1 or fractional T-1 leased line services can be mapped into an ATM ring simply by following the ATM Forum standard for circuit
emulation service (CES). In comparison, IP has no definition of how to transport such services. Since CES defines parameters associated with the
quality of services such as latency, the performance expected for the same service in traditional Sonet can be achieved by simply specifying this service
parameter.
The standards mapping also assures that this service can be transported reliably over the ATM backbone since it follows the same rules, and the
hand-off to the conventional circuit switched elements is also easily achieved. By using inherent multiplexing capability of ATM, T-1s and fractions of
T-1s can be "groomed" right within the ring, eliminating the need for external 0/1 and 1/3 digital cross-connects (DCS). This capability can lower the
provisioning cost to a level where it is cheaper to deliver these services over ATM rings than over traditional Sonet.
Similar is the situation with transparent LAN service (TLS) defined as a point-to-point connection between two LANs. Earlier we described the
complexity with which this service is delivered today. With ATM, Ethernet can be mapped directly into the ring using the IETF-defined RFC 1483
standard. This approach eliminates a separate router, which is overkill for delivering this simple bridged service. In addition, only the bandwidth
required for this service is transported over the ring--eliminating bandwidth waste common with traditional Sonet implementation.
Because this solution is "transmission like" in its provisioning simplicity, a carrier's transmission organizations can easily become the suppliers of data
service solutions. When the same approach is used for Internet access, traffic from multiple Ethernets can be groomed to a single hand-off port,
dramatically lowering the cost of backbone infrastructure (Figure 2). The end result is lower provisioning and operations cost, and a services rollout
capability that is much faster than today.
In the case of frame relay service access, today it is delivered in fixed bandwidth pipes (typically T-1 or fractional T-1)--first to frame relay switches,
which then perform statistical multiplexing and pass "groomed" traffic to the ATM backbone. In many cases, a digital cross-connect is also needed to
groom T-1s and fractional T-1s into high-density interfaces (like a channelized DS-3) to save valuable port space in frame relay switches.
With ATM rings, frame relay can be mapped directly into ATM, statistically multiplexed in the transmission ring, and delivered directly into the ATM
backbone. No more cross-connects or frame relay switches. It is easy to see how this simplification can make service creation a much more profitable
affair.
------------------------ cut ------------
Pay attention to where it said about not needing digital cross-connects, and routers. TLAB needs CIENA more than CIENA needs TLAB.
|
