What's Cisco's WAN strategy?It's anyone's guess

SAN JOSE - Now that Cisco has killed off its ATM switch for the core of enterprise and service provider networks, the company's plans for the wide-area market are unclear.

Cisco is banking on increased sales of WAN switching gear to enterprises and service providers to drive the company's growth. But last week's news that Cisco discontinued development of its core WAN switch and delayed another enterprise switch for a year indicates that success for Cisco in this market may be more challenging than expected (NW, May 3, page 1).

Analysts, competitors and other Cisco watchers say the company's long-term WAN switching strategy for enterprises and service providers still hasn't jelled three years after Cisco's $4 billion acquisition of StrataCom. They say Cisco has shipped only one new WAN switching platform - the MGX 8850 IP/ATM edge switch - since the company acquired StrataCom. And that offering is available in limited volume.

Others products are merely upgrades of existing StrataCom architectures. And Cisco has lost market share in frame relay WAN switching since acquiring StrataCom, according to Vertical Systems Group of Dedham, Mass. (see graphic).

"In frame switching Cisco has decreased because [Ascend's Cascade switches] got developed and rolled out," says Rosemary Cochran, principal analyst at Vertical Systems Group.

Cisco rebuffed repeated requests from Network World this week for interviews with company executives on the state of its WAN switching business. Cisco did, however, respond to queries from other publications regarding this topic as those publications contacted Network World for commentary.

The death of the 20G bit/sec TGX 8750 seems to leave Cisco without a core IP/ATM WAN switch to challenge offerings from Ascend, Newbridge Networks and Nortel Networks, and with a gaping hole in its "end-to-end" voice/data system story. From a short-term revenue standpoint, that may not be disastrous because Cisco has said in the past that the revenue potential at the edge of the network is 15 times that of the core.

But longer term, a lack of presence in the core may mean a lack of account control and the incremental revenue increases that control provides.

In any event, the TGX 8750 was supposed to serve as proof of the synergy between Cisco and StrataCom - that IP from Cisco could be "married" to ATM from StrataCom to provide the best of packet and cell switching for enterprises and service providers.

Cisco now says its MGX 8850 edge switch, which scales from 1.2G bit/sec to 45G bit/sec, can slide into the core. Indeed, Cisco customer Sprint always intended to use the MGX 8850 as its core switch for its Integrated On-Demand Network, a Sprint spokesman says.

But it is still unclear whether the MGX 8850 is a tactical or strategic platform for the core.

Cisco competitors say at 1.2G bit/sec, the MGX 8850 currently lacks the horsepower for edge duty, let alone core. They say that by the time the MGX 8850 scales to 45G bit/sec - which they believe to be mid-2000 - the switch will have already been surpassed by other products.

Cisco's strategic IP/ATM switch for the WAN core, competitors say, is a 120G to 190G bit/sec platform under development, code-named Jupiter. They expect Jupiter to ship in late 2000.

There's also always the possibility that Cisco could acquire its way back into the WAN core by snapping up one of the gigabit/terabit router start-ups. Juniper Networks and Avici Systems may be off-limits, given that several Cisco rivals have equity stakes in Juniper and Nortel owns 20% of Avici.

But acquiring a router start-up at this stage would be a humbling experience for Cisco, observers say. That move would signal that the $4 billion StrataCom acquisition did not pan out, and that Cisco, the worldwide leader in routers, did not have the wherewithal to develop a high-speed switching router for the WAN core - that marries IP and ATM - in a timely fashion.

Whether Cisco acquires a company or Jupiter emerges, analysts say Cisco to date has been sending mixed messages regarding its strategic technology for the WAN core: IP packets or ATM cells. Though Cisco claims to be "technology agnostic" - having no preference of one over another as long as it offers what the customer wants - the company has actually been downplaying the significance of ATM in next-generation data optimized networks.

"Without a core switch, Cisco is going to continue to try and marginalize ATM, except at the edge," says Craig Johnson of The PITA Group in Portland, Ore. "It's to the company's advantage to do such a thing and say that routers are where the intelligence is."

"Cisco is still schizophrenic with regard to ATM and routing," says Tom Nolle, president of consultancy CIMI Corp. in Voorhees, N.J. "The strategies that it's talking about are not consistent with its product positions. The service providers are suspicious of people who they think are maybe talking out of both sides of their mouths."

Indeed, Cisco is stating three reasons for killing the TGX 8750, one of which is slow demand for OC-48 ATM in the WAN core, an assertion Cisco competitors and analysts say is ridiculous.

"I don't agree with that," Vertical's Cochran says. "Certainly demand hasn't decreased from a year ago; if anything, it's increased."

Cisco has underscored the OC-48 packet-over-SONET features of its 12000 GSR router as key to the product's selection by service providers Frontier, France Telecom, IXC, Swisscom and Enron.

Another reason Cisco gave for discontinuing the TGX 8750 is that the company could not build a single product at prices that enterprise network customers and service providers expect.

Cisco's third reason, according to sources, is that the 20G bit/sec TGX 8750 no longer makes sense for the core when the MGX 8850 scales to 45G bit/sec.

Observers say Cisco's explanations for the demise of its core IP/ATM switch indicate that the company's WAN switching strategy is just as scattershot - even three years after the StrataCom acquisition. This, along with the release of only one new WAN platform in three years and the loss of frame relay market share, signals that Cisco has so far benefited little from StrataCom, and vice versa.

The only apparent gain from StrataCom is that Cisco bought its way into the AT&T and WorldCom public frame relay networks.

"They've renamed a bunch of products, but I haven't seen much more than that," says Bob Bellman, president of BrookTrail Research in Natick, Mass. "Nothing exciting, anyway."