By JIM DUFFY
Network World, 07/24/00
When announced in June 1998, the merger of Nortel and Bay Networks was hailed by officials of both
companies as a metamorphosis in the industry.
The move was an admission by an old guard telecommunications equipment giant that the future of
networking is in data - even for voice. The deal remains the most aggressive move to date by a telco
stalwart to transform itself into an IP packet-switching powerhouse virtually overnight.
"Data has a growth rate of 30% to 40% per year, and it's driven by the Web," Nortel CEO John Roth said
in 1998, explaining the rationale behind the $9.1 billion union. "The IP component is growing more like
70% per year. That's really the segment we're focusing on."
But Nortel's IP market share, specifically in routers and Layer 3 switches, has done anything but grow.
The company confirmed last week that after a significant delay, it will not ship the 240G bit/sec Versalar
25000 router for the Internet core as a stand-alone product. It's been "folded" into the OPTera Packet
Core, the cornerstone of Nortel's product portfolio for merging packet and optical networks.
In the enterprise, Nortel has upheld Bay's tradition of losing market share to Cisco in high-end routers.
Nortel's share in this market fell from 15.7% in 1998 to 11.2% in 1999, while Cisco's share increased
5.5%, according to The Dell'Oro Group in Portola Valley, Calif.
In Layer 3 switching, which was another Bay stronghold, Nortel's dive has been even more dramatic.
Nortel lost its market leadership and more than half of its market share between 1998 and 1999, according
to Dell'Oro - from 37.8% to 17%, a drop of almost 21 percentage points. Nortel also delayed the
introduction of its next-generation Layer 3 switch by three quarters.
Cisco again was the chief beneficiary of Nortel's precipitous fall, gaining 26.2 percentage points between
1998 and 1999, according to Dell'Oro.
As Nortel marks the second anniversary of its Bay acquisition next month, the telecom giant seems to be
anything but an IP company. Indeed, Nortel appears to be losing its grip on IP and having little to show for
its $9.1 billion investment.
"In this space they have missed it," says Frank Dzubeck, president of consultancy Communications
Network Architects in Washington, D.C., referring to IP routing and switching. "They are falling behind
tremendously in everything but optics."
How did it happen?
Along with Cisco, 3Com and Cabletron, Bay was one of the prestigious "Big Four" in enterprise data
networking, a handful of multibillion-dollar companies that accounted for more than 80% of the hub, switch
and router market. Bay was created by the 1994 marriage of two leading internetworking vendors:
SynOptics in shared-media Ethernet hubs and Layer 2 switches, and Wellfleet, the No. 2 router vendor -
albeit a distant No. 2 - behind Cisco.
Bay's fortunes in routing began to sag long before Nortel entered the picture. The company steadily lost
market share to Cisco for most of the 1990s, and lost half its share between 1996 and 1999, according to
Dell'Oro.
Enter the Layer 3 switches
Part of that decline had to do with the emergence of Layer 3 switches, a cost-effective, high-performance
hardware-based collapsed backbone alternative to routers. Bay, though, was quick to pounce on the
opportunity by snatching up Gigabit Ethernet start-up Rapid City in June 1997, and Bay's dexterity quickly
earned it the leadership position in this burgeoning market.
But Nortel's acquisition of Bay changed all that. Bay and Rapid City engineers who were used to working
at a nimble, fast-moving company with a start-up's pace, fled the buttoned-down, bureaucratic Nortel. This,
along with the distraction of integrating its $9.1 billion prize into the corporate fold, crippled Nortel's ability
to maintain - let alone build on - its market leadership in Layer 3 switching.
The result was the plunge in Layer 3 market share from 1998 to 1999, and Cisco's dramatic gain.
"I don't think [the Bay acquisition] was a bust, but it seems like the traditional voice companies are having
a harder time transitioning to packet-based than companies like the Ciscos are transitioning to voice,"
says Eric Thompson, principal analyst at Dataquest. "Nortel really needs to get in gear and really start
touting its wares and delivering product that moves [the company] into this next-generation network."
The Layer 3 switch numbers don't tell the whole story, says Arun Jain, director of IP infrastructure
solutions at Nortel, who joined the company with the Bay acquisition.
"Our market share percentage has gone down, but it was obscenely high because we were the first one to
come out," he says. "Over time, although our percentage has come down, our business has grown
significantly."
That it has. Even though Nortel lost more than 50% of its share of the Layer 3 switch market between
1998 and 1999, revenue rose 50%, from $234.3 million to $351.1 million, according to Dell'Oro. But
virtually everyone grew by at least that much as the worldwide market for Layer 3 switches ballooned from
$620.2 million in 1998 to $2 billion last year.
Even 3Com, which recently discontinued its Layer 3 switches for the core of large companies, increased
its revenue by about 8.5%.
Nortel did not help itself last year - or the first half of this year - by delaying the release of its
next-generation Layer 3 switch by nine months. The Passport 8600 is the follow-up to the Accelar 1000
line, which was the first-generation Rapid City product.
The Passport 8600, initially intended for an October 1999 debut, began shipping in June. Nortel officials
said they delayed the switch to get it right. Widener University in Chester, Pa., which recently awarded
Nortel a multimillion-dollar contract to replace a 3Com network with 8600s, concurs with that (www.
nwfusion.com, DocFinder 8845).
"We expect that business to start growing significantly again," Jain says.
Routing may be another story. As the market for enterprise routers was evaporating for everyone but Cisco
in the mid-1990s, a robust demand for routers was germinating in the service provider market, due to the
commercialization and growth of the Internet and the deregulation of the telecom industry in 1996. These
two events served as the impetus for Nortel's acquisition of Bay.
But the cancellation of the Versalar 25000 - which was based on a product obtained from Bay - confirms
that Nortel has gotten off on the wrong foot in service provider routing. Nortel scuttled plans last year to
resell a terabit router from start-up Avici Systems, in which it has a 20% stake.
Nortel may be a player in Internet core routing, but not in Internet core routers.
"When the Versalar 25000 was going to come out to market toward the end of this year, we didn't think it
would get us a leading position," Jain says, explaining the rationale for folding the product. The Versalar
25000 was supposed to ship late last year. "It had a proposition that was very strong, but we would have
been at best a No. 3 player."
According to Dell'Oro, Nortel had a minuscule 1.3% of the nearly $694 million market for WAN routers in
1999, due largely to the Versalar 15000 edge router - another product with Bay roots - and the Shasta
5000 Broadband Service Node, which was obtained via acquisition. Nortel is well behind Cisco's 78.7%
share and Juniper Networks at 14.8%.
Indeed, observers consider WAN routing to be a two-horse race between Cisco and Juniper even though
Dell'Oro currently counts about 23 players - many of them start-ups - in the market.
Nortel merged Versalar 25000 processor and software technology into the OPTera Packet Core, a switch
fabric that transports IP, ATM and SONET traffic, and serves as the interface between the packet network
and the optical network. OPTera Packet Core is a component of Nortel's OPTera Packet Solution (OPS),
a five-product package designed to meld packet and optical networks into a high-performance multiservice
infrastructure for the Internet core.
OPS will ship next year, Jain says.
"A lot of the key technologies in terms of both software and processors from the Versalar are being
merged in the OPTera Packet Core," Jain says. "We have an IP/optical play where we believe we can get
a leadership position."
Analysts had high hopes for the Versalar 25000.
"It is an integral part of their whole OPTera proposition," says Ron Westfall, an analyst at CurrentAnalysis
in Sterling, Va. "Without it, it's a gaping hole in their ability to execute on that particular proposition."
Jain says the Versalar 25000 probably wasn't as integral as people thought, or as Nortel led them to
believe.
"It was an edge or peripheral device to OPTera Packet Core," he says. "It was always intended that that
edge or peripheral device could always be the Versalar 25000, or some other router from some other
vendor as well. At this point, we're not going to use the Versalar 25000 as the peripheral or the edge: We
will use a product from vendor A or B."
Vendor A is clearly Juniper. As Nortel extended the delay in the Versalar 25000, the company and Juniper
signed a "memorandum of understanding" last month to jointly market, plan and implement optical core
routing infrastructures for service providers. The agreement involves assurances that the two companies
will foster interoperability between Nortel's OPTera optical switching and transport systems and Juniper's
M-series Internet core routers.
Jain acknowledges that Nortel's progress in IP routing since the Bay acquisition has been disappointing.
But Bay instantly gave Nortel leadership positions in VPNs and cable modems, and also gave Nortel
7,000 to 8,000 people trained in IP data networking, he says.
"The only place where we have had not as much success as we originally hoped is in routing," Jain says.
"Part of that is [because] competition is fierce in that space."
The other part is Nortel's focus on leadership in optical networking, which is fine with some of its
customers. Pathnet, a "carrier's carrier" in Reston, Va., says the company has some optical IP trials
under way that will dispel any notion of Nortel not being an IP company.
Nortel has a method for binding Multi-protocol Label Switching, which is used for engineering traffic
through IP and ATM networks, directly to dense wave division multiplexing gear that will drive quality of
service down to the lambda level, and may be years ahead of competitors, says Gerry Sharp, Pathnet's
chief technology officer. He says Nortel's technique alleviates the need to convert wavelengths to electrical
impulses and back again.
"They will begin to attract more attention in this industry from an IP perspective," Sharp says of the Nortel
trials. "Their IP products - this product - is an optical-to-optical solution: It is not optical-electrical-optical.
This will bring them back onto the table for IP."
But until Nortel can generate revenue and significant market share with this new optical IP product,
observers remain skeptical of the company's progress in IP since acquiring Bay.
"Roth basically said he was going to transform his company into an IP company. There is no evidence
that that transformation has ever occurred," Communications Network Architects' Dzubeck says.
ragingbull.altavista.com |
