Mike,
The rest of the article...
Ascendcontinues to integrate IP and routing support as a feature on its ATM switches. The company will focus IP enhancements on the CBX 500 and B-STDX 9000 edge switches, leaving the core GX-550 platform as an ATM-centric product, at least for now. We believe this strategy appropriately plays to the company's strength as a market leader in ATM, and Ascend firmly believes that most carriers will deploy hybrid IP and ATM networks. Nonetheless, Ascend could be at a slight disadvantage when competing for “green field” customers that lack a proclivity towards ATM, in our view.
Cisco is the market leader in every router category, and holds roughly 70% of the router market through the first nine months of 1998. Many suspect that the company's market share is even higher among service providers, and approaches 80%. Cisco does about $1.1-1.2 billion per quarter in total router sales, which presently represents about 40-45% of revenue. This percentage has been declining steadily as the company diversifies its business into faster-growing areas. Within the service provider community, Cisco's venerable 7500 series products are positioned as edge routers, with the newer GSR 12000 routers targeted at the Internet core. Cisco recently announced it had reached the 1,000 GSR unit milestone, and can claim as customers Internet mainstays that include Sprint, Qwest, Frontier, Enron, GTE, MCI and AOL.
We believe Cisco's dominance is a direct result of its software expertise. Products have come and gone that possess better specs than Cisco's, and the company usually lags the market in raw performance. Yet, no one has been able to displace Cisco from the top spot. Cisco is often first-to-market with new software features and, in our view, the company's Internetworking Operating System (IOS) software has become the de facto standard for Internet routing. In addition, Cisco holds key positions in many standards organizations, and is extremely influential in setting new Internet and protocol standards. One important example is with MPLS, wherein Cisco is touting its own version of the specification, called Tag switching. Although the company plans to endorse and support MPLS, it inevitably will add its own proprietary flavoring above and beyond the standard implementation. The end game is to provide carriers with a solution more valuable than a pure standards-based approach, which in turn, locks the customer into a Cisco-centric migration path.
Although Cisco continues to enjoy the leading market share position, it remains vigilantly paranoid of potential competitors. The company recently announced the Cisco 7576 to protect its territory in the edge router market. The 7576 essentially doubles the density of existing 7507 routers without an increase in footprint. We also expect Cisco to introduce additional products in this space in mid-1999, potentially based on scaled-down versions of the GSR 12000. Major improvements would likely be in packet per second performance, in which the 7500 has respectable but not exemplary metrics, and greater support for OC-12 and OC-48 line speeds, in which the 7500 could be competitively disadvantaged.
In addition, Cisco has split the development of 12000 from the rest of the router product line; it is now part of a relatively new optical internetworking business unit. We believe this organizational structure portends continued development and an expanding product offering with respect to optical networking. One example is the company's recent acquisition of Skystone, which provides enhanced optical management features that Cisco plans to integrate into the 12000. To date, the Company has signed deals to supply equipment for IP over WDM architectures to Enron, Sprint and Frontier, with several more unidentified prospects in the pipeline.
Lucent traditionally has not been a player in the data networking market, but is certainly a powerhouse in telecom equipment. As those two worlds converge, Lucent is likely to increase its presence in data networking through acquisition and internal development. In May 1998, the Company announced its first IP routing products for service providers, the PacketStar 6400 IP switch family. Based on the product specifications it could be a contender in the carrier-class switch/router arena, in our view. There are 3 PacketStar models that range in performance from 1 million to 32 million packets per second forwarding. The flagship 6416 has 64 Gbps of switching capacity and can support up to 16 OC-48 interfaces, which places it in the hunt with comparable competitive products. Concurrent with the product release, MCI was announced as the first customer to trial the router. We view this as particularly significant event for two reasons. First, MCI, soon to be Cable & Wireless, is one of the largest providers of national Internet backbone services in the U.S. and is influential in determining the evolution of the Internet's infrastructure. Second, MCI historically has shunned Lucent's products because of Lucent's ties to AT&T, MCI's primary competitor in residential and business voice services.
Based on Lucent's historic aptitude in engineering telecom networks, we would expect the company to emphasize its bandwidth management, quality-of-service and reliability features, which are critical functions. In addition, Lucent's voice expertise is unsurpassed and few established players match the company's voice reputation, much less newly formed start-ups. We also believe Lucent will leverage its WaveStar WDM and optical networking product portfolio for customers stressing direct IP to optical layer integration. Although Cisco is working with Ciena in this area, and Northern Telecom also has some optical equipment, Lucent is the only company that currently is shipping both WDM and carrier-class switch/router products.
Despite Lucent's incredibly broad product arsenal, strategic account relationships, and first-class customer service organization, the company has yet to have a serious impact in data networking. Lucent has become a more active acquirer over the past 18 months, shoring up specific holes in its data offerings. We would not be surprised to see the company add to its router portfolio through acquisition if the PacketStar line proves unsuccessful.
Juniper perhaps has received the most external attention of the carrier-class router start-ups despite a low profile, somewhat secretive, approach to market. This attention is likely a direct result of the corporate investors that hold minority interests in Juniper, which include 3Com, Ericsson, Siemens/Newbridge, Northern Telecom, Lucent and UUNet, and in part because Juniper's team contains a high concentration of former Cisco employees. The company's first product, the M-40 router, began shipping in September, and Juniper has already announced initial customer wins at UUnet, Frontier, Verio and MCI. The base price for the M40 is $55,000, with individual interfaces starting at $25,000.
The M40's mission is to become the foundation of the optical Internet, providing scalability in terms both of raw throughput and software richness and control. To achieve this goal Juniper believes routers must support the transition to higher-speed interfaces (e.g., OC-12 and OC-48), and deliver more than just best-effort service. The company has shunned microprocessor-based or assisted system architectures, which it believes are not capable of delivering the performance required of next-generation routers, in favor of an ASIC-based forwarding engine.
When it comes to the debate over Internet architecture, we would not necessarily place Juniper in the anti-ATM camp, but the Company is definitely pro-routing. Juniper feels that ATM cores have several disadvantages including the ATM cell tax, scalability of ATM circuits in a meshed environment, less resiliency and greater management complexity arising from overlay networks. Nonetheless, ATM offers better traffic engineering and efficiency than do routed cores. The company's strategy is to use MPLS to bring ATM's advantages to a routed network and to provide ISPs with a mechanism for traffic engineering and differentiated service levels.
To our knowledge, Nexabit's NX64000 terabit switch/router has the greatest claimed capacity of any announced product. The company touts that its flagship platform can route 6.4 Terabytes per second in a single chassis. This level of capacity is intended to convert the raw bandwidth of the Internet into usable infrastructure, to support new services, and to enable voice and data infrastructure convergence. The company dismisses competing architectures as inadequate for keeping pace with Internet growth. Nexabit claims to have developed a new hardware architecture and has filed for no less than 14 patents related to that effort. Unfortunately, due to competitive reasons, technology details are scarce.
What we do know is that the NX64000 can natively route IP packets and switch ATM cells, and also supports frame relay and MPLS. One of the router's unique features is called IP constant bit rate (CBR), which matches ATM's class of service levels in an IP environment. CBR is often used when transporting voice over ATM, and IP CBR could be beneficial for carriers implementing voice over IP.
From a speeds and feeds perspective, Nexabit believes that OC-48 interfaces will be the “sweet spot” for carrier-class routers through 1H 1999. However, the company has no plans of stopping at OC-48, and built the NX64000 with scalability in mind. Accordingly, Nexabit talks about supporting up to OC-3072, an unfathomable 160 Gbps, in future iterations of its router. The initial release, which is expected to ship in early 1999, supports up to 16 OC-192 interfaces in a 16-slot chassis. In addition to blinding performance, Nexabit has engineering several attractive reliability features into both software and hardware components. The most interesting is a feature called Software HotSwap, which allows customers to change the router's software image in a piecemeal fashion, without taking it off-line for the update. The NX6400 also permits individual line cards to be rebooted without impacting the overall system.
All of this performance and functionality does not come cheap, and entry-level pricing is
purported to start around $500,000 and go upward from there based on configuration
|
