Urban legends often become accepted as fact even though they aren’t true. An urban legend is based on a compelling idea or concept that seems plausible. They are often based on a factual event or concept, and have become even more prevalent with the advent of the Internet and e-mail. These myths are spread by publications and colleagues--authoritative sources that add to their credibility. In truth, many are no more than fiction.
But which do you recognize as myths? Which are you or your organization acting on, and which are your competitors following? The difference to your business can be immense, and it can mean the difference between success or failure. Here, we’ll examine five common networking myths and separate fact from fiction.
Myth #1: ATM is Dead
This myth has gained wide currency in the last few years for a number of reasons. First, ATM didn’t win the desktop war, therefore its perceived overall value was deflated. Second, many have come to believe that ATM can’t scale to high speeds, diminishing its ability to handle the blistering bandwidth growth brought on by the Internet. And finally, it’s widely believed that the “cell tax” killed ATM––that the cell header overhead rendered ATM inefficient.
The reality of ATM, however, is different. While ATM to the desktop is dead, it’s alive and well in the access layer and core of service provider networks. In fact, according to industry consultants, the Yankee Group (www.yankeegroup.com), ATM switch revenues continue to grow at 36 percent per year and ATM switches will generate worldwide revenues in 2003 of $7 billion. This growth is being driven by factors such as widespread deployment of DSL and wireless networks that use ATM cells for transport, ATM multiservice access, and TDM network replacement. In addition, MPLS (Multiprotocol Label Switching) has given ATM a new lease on life by vastly simplifying the provisioning of IP services on an ATM network (see “Comparing IP over ATM to IP+ATM” diagram, page xx). MPLS is an emerging Internet Engineering Task Force (IETF) standard based on Cisco’s Tag Switching. MPLS uses a label-based forwarding paradigm to enable carriers to deliver a highly scalable foundation for the delivery of value-added IP business services in an IP or ATM network.
Figure 1 – Comparing “IP over ATM” to “IP+ATM”
With the traditional “IP over ATM” model, a complex IP to ATM protocol and addressing translation must be performed. In contrast, an IP+ATM model allows IP addressing and routing to be directly translated into ATM forwarding decisions. Once a packet has been labeled at the edge of the network, core MPLS switches need routing for the control path only –– and not in the data path –– improving performance and significantly reducing complexity. With MPLS, IP+ATM switches make the same forwarding decisions as a router, based on information from the same IP routing protocols. By “IP-enabling” the ATM switch, service providers can easily leverage their ATM networks in deploying new IP services.
Incumbent service providers such as AT&T Corp. (www.att.com) are IP-enabling their ATM and frame relay networks with MPLS to deliver new services such as IP Enabled Frame Relay (IPFR), a connectionless IP VPN service. Competitive carriers such as Caprock are building new MPLS-based IP+ATM networks that deliver revenue from transport services today as well as revenue and profit growth from value-added IP services. These New World IP+ATM networks reduce the cost of traditional services and provide new services with faster time to market.
While the ATM myth holds that ATM won’t scale to high speeds, in fact, there are ATM interfaces running at OC-48c/STM-16 in lab trials today and OC-192/STM-64 is in development.
While the cell header overhead is a consideration for some applications, such as long-haul intercontinental links, the cell header allows ATM to deliver a wide range of services: from IP to circuit emulation. ATM is highly effective for migrating TDM services such as private (or leased) lines to packet-based networks, allowing service providers to cap their investments in time division multiplexing (TDM) and strategically invest in packet networks for future growth and profitability.
Myth #2: IP Can’t do Real QoS
The second widespread myth is that IP can’t do real quality of service (QoS) and, therefore, is not ready for prime time. But the facts tell a different story. Today, BCE Nexxia (www.bcenexxia.com), Global Crossing Ltd. (www.globalcrossing.com), and Digex Inc. (www.digex.com) are only a handful of the service providers that are deploying IP QoS mechanisms for Internet access, VPNs, and VoIP services.
IP QoS may seem complex since it involves many different technologies. While circuit-oriented QoS for ATM or frame relay is easy to understand, it lacks the flexibility that is critical in today’s networks. One way to understand how IP QoS can be applied in an any-to-any network is described by the following model (see “IP QoS Model” diagram).
Figure 2 –IP QoS Model
At the top layer of the above model, Cisco’s Committed Access Rate (CAR) feature provides IP-level bandwidth management and packet classification. CAR uses IP precedence bits to implement network QoS policy. It also limits (or reserves) access bandwidth based on a wide range of criteria such as IP destination or application.
The middle layer of the model highlights how service classes are defined in the service provider’s network. For example, three classes could be defined: best effort, guaranteed latency, and guaranteed latency and delivery. All traffic would be classified into one of these classes. Queue management techniques such as Weighted Random Early Detection (WRED) and Weighted Fair Queuing (WFQ) could be used to implement and enforce these classes. Packet classification technologies like CAR and the use of service classes have recently been standardized in the IETF as part of the Differentiated Services (DiffServ) standard. Contrary to some recent misinformation, DiffServ and MPLS work well in conjunction with each other, and are not competing technologies.
Traffic engineering techniques operate at the lowest layer in this model to make the greatest possible bandwidth available for traffic of all service classes. Cisco’s MPLS Traffic Engineering does this by adjusting routing in the network to utilize any spare capacity. As part of this process, MPLS Traffic Engineering reduces congestion by routing traffic away from congested links whenever possible. MPLS Traffic Engineering supports IP QoS by making the greatest possible uncongested bandwidth available.
While some rightly claim that using a separate signaling mechanism like RSVP for each IP flow limits scalability, the reality is that RSVP is not required for each IP flow. For example, in a network with 1,000,000 hosts and 1,000 routers or MPLS switches, RSVP might not be scaleable if it were used for the individual flows between the 1,000,000 hosts. However, RSVP is quite scalable when used only for the aggregated traffic streams between the 1,000 routers and switches. This is the role RSVP plays in MPLS Traffic Engineering.
Myth #3: Optical Networking is Only for IP
A number of myths have sprung up around IP, ATM and optical internetworking –– the technology that combines gigabit and terabit internetworking (routers and switches) with current and emerging optical transport technologies to transport data over optical wavelengths. The most common myth is that optical internetworking is for IP only.
The fact is that IP and ATM are complementary technologies and are both supported on optical internetworks.
“The Role of IP and ATM in Optical Internetworking” diagram on page xx highlights the role that both IP and ATM play in optical internetworks. At the bottom, the optical fiber layer maximizes the investment in the transport infrastructure by delivering the dramatic capacity now being demanded on public networks. SONET/SDH is the predominant interface today between the fiber and the higher layer protocols of ATM or IP. SONET provides operational continuity and key restoration capabilities such as Automatic Protection Switching (APS) as networks transition from old world TDM to New World packet networks. ATM, which can interface to the fiber directly through transponders such as wavelength division multiplexers (WDM) or through SONET, provides the connectivity for legacy services such as frame relay, ATM, and private lines. At the highest layer, and also interfacing directly to fiber through WDM or through SONET, IP delivers the new Internet IP-based applications, such as Internet access, VPNs or web hosting. MPLS is the glue that brings IP and ATM together.
Figure 3 – The role of IP and ATM in optical internetworking
Myth #4: MPLS is only for ATM
This myth holds that MPLS works only for ATM. How has this come about? It has most likely gained credence because MPLS brings so much value to ATM-based carrier networks. If you’re a service provider with a frame relay or ATM network in a world that is quickly moving to IP, your first imperative is to IP-enable your network. MPLS is a powerful and innovative solution that enables these ATM-based carriers to support revenue-generating services such as frame relay and private-line replacement today, while providing a seamless path to new IP services––the drivers of tomorrow’s revenues and profits.
The reality is that MPLS provides a foundation for IP service delivery for networks based on ATM switches or IP routers. MPLS provides a transport-independent service creation layer that enables easy grouping of users and services with MPLS-based VPNs (see “MPLS: The Basis for a Transport-Independent Service Delivery Infrastructure” diagram, page xx). It delivers IP VPNs with the QoS and privacy of ATM and the any-to-any connectivity of IP, without the administrative costs and scalability limitations of tunneling or encryption. What’s more, MPLS provides universal access: a single MPLS-based VPN can work across a DSL, ATM, and dial network, providing each connection with access to all services and applications, including voice, content hosting, net commerce or video collaboration. With VPNs based on tunnels or virtual circuits, applications and content are forced outside the network, limiting the service provider’s involvement and control; with MPLS-based VPNs, service providers can add value though applications and content inside their network.
Figure 4 – MPLS: The Basis for a Transport-Independent Service Delivery Infrastructure
Another part of the MPLS myth is that MPLS doesn’t add value to IP networks. But this, too, is incorrect as MPLS combines the benefits of ATM with those of IP––for ATM networks and IP networks. Through labeling, MPLS allows carriers to set up deterministic, switched paths––common with ATM––with the flexible any-to-any capability needed with today’s applications. In effect, IP-based carriers gain the ability to do both traffic engineering (because MPLS-based label-switched paths are deterministic) and the ability to offer high QoS for time-critical traffic by assigning labels that deliver the specific traffic quickly with low latency. This is the scenario being implemented by Toronto-based BCE Nexxia, one of Canada’s leading carriers. BCE Nexxia offers customers its fully managed Nexxia.IP VPN™ service Enterprise using MPLS across its carrier-class, coast-to-coast OC-192 SONET/dense wavelength-division multiplexing (DWDM) IP network.
Myth #5: IP and ATM are Competing Technologies
The final myth is that IP and ATM are competitors and that IP beat ATM in the “protocol war.” The reality, however, is that the battle is packets versus circuits, not IP versus ATM. New World communications is based on packets, not just IP. IP and ATM are complementary packet technology options that provide different services and capabilities. Carriers should choose one or the other (or both) based on business and service decisions, not technology factors. The right answer might be an integrated IP+ATM network, separate IP and ATM networks, or IP and IP+ATM networks linked together with MPLS.
AT&T’s IPFR service is an excellent example of how a carrier IP-enabled its ATM network and deployed an IP VPN service that complements its existing frame and ATM services.
The IPFR network allows customers of AT&T’s frame relay network to buy a frame PVC that connects to a connectionless IP VPN inside the AT&T network. The IP VPN provides the any-to-any connectivity that their new IP business applications require and at a substantially lower cost, since only one PVC is required per site (see “ AT&T’s IP-Enabled Frame Relay Service Compliments Existing Frame Relay Services” diagram, page xx). With the traditional layer two point-to-point solution, costs rise exponentially for both the user and the carrier as more connections are added (requiring additional PVCs for a partial or full connectivity mesh).
AT&T’s customers benefit from improved performance and simplified management. Migration is simplified since they can continue to use frame relay to connect to the IPFR service, and the IP VPN network is just as secure as an individual frame relay connection.
Figure 5 – AT&T’s IP-Enabled Frame Relay Service Compliments Existing Frame Relay Services
Conclusion
In networking, issues are often cast as one technology versus another, such as IP versus ATM. But the truth is often more subtle and complementary.
With IP and ATM, the issue is not one technology versus another. The real issues revolve around the business drivers. What technologies will support the business goals of generating revenues from today’s ATM, frame relay and private line services and provide a path to packet networks that will enable future revenue and profit growth? Service providers must focus on the services and business plan first, design an architecture to support these revenue and service goals, and then decide on the right products and technologies to solve their business problem. Once evaluated in this light, carriers of all kinds are finding that both ATM and IP are critical technologies for securing profits today and evolving to tomorrow’s lucrative IP services.
Rob Redford is director of marketing at Cisco Systems Inc. (www.cisco.com). He can be reached at (408) 527-3465.
