Fast router feast
CEOS from five high-speed router start-ups
chow down on convergence, QoS and other hot
topics.
By Jeff Caruso and Bob
Brown
Network World, 11/02/98
Atlanta - The high-speed
Internet router start-ups
haven't conquered anything
yet, but their dreams sure are
big.
That's why having lunch at a
popular restaurant here called
Veni, Vidi, Vici ("I came, I
saw, I conquered") with the
CEOs from five of the
start-ups seemed a bit
premature. It was also
strangely appropriate, because
these are the leaders who
came to the Internet and saw
the need for boxes that route
faster than anything the world
has ever seen. But actually
building these gigabit- and
terabit-speed routers and
conquering the market - well, that may be another
story.
For now, at least, these guys are stoked, just like the
Gigabit Ethernet start-up CEOs we wined and dined
during several such get-togethers over the past two
years. While the router companies have no plans to
sell their wares to enterprise customers, their products
promise to have a big impact on the types of services
users can expect to buy from carriers and ISPs down
the road.
We stole the Internet router executives away from the
crowds of NetWorld+Interop 98 last month to chat
with them about how they plan to change the Internet
as we know it. Around the lunch table were five of the
Internet's brightest: David Bernstein of Pluris, Mukesh
Chatter of Nexabit Networks, Ashraf Dahod of
NetCore Systems, Mike Grady of Argon Networks
and Surya Panditi of Avici Systems.
(Yes, Juniper Networks CEO and Interop keynote
speaker Scott Kriens was conspicuously absent from
our table. We did invite him because Juniper and its
$62 million in venture capital have generated much of
the buzz around superfast routers, but we have to
admit our invites went out a bit late.)
A quick survey around the table revealed that the five
vendors represented had garnered a whopping $185
million on their own, with funding ranging from $14
million for NetCore to $72 million for Avici. Even so,
we insisted on paying for lunch.
The start-ups' venture funding may seem like a lot of
money, but these vendors feel justified in asking for
the cash. The CEOs said they need that kind of
capital to build products required by service providers
to handle ever-increasing traffic loads. "[Big
investments in Internet router companies] are a
recognition by the venture community that this is a
difficult problem, and you don't solve it for the
traditional 7 or 8 million bucks," said Grady, who in a
previous life helped launch Stratus Computer.
It was readily apparent that these guys aren't just
trying to build a new core to the Internet. They're
trying to pave the way for a new Internet, one that's
reliable and versatile enough to carry voice and video
as well as data. They realize that enterprises aren't
going to trust carriers to transport their most critical
traffic - unless the Internet router vendors can make
devices as reliable as central office voice switches are
today.
"The requirements for availability, reliability and
serviceability - all the things that matter to carriers -
are what will enable enterprises to move their traffic to
carrier services such as VPNs," said Panditi, who
complimented us on our choice of a restaurant he
likes to call Veni, Vidi, Avici.
Another quick survey around the table showed that
none of these companies has yet shipped a product.
The earliest of their high-speed Internet backbone
routers will start trials by year-end, and others won't
ship until well into 1999.
We recognized that by inviting our guests to lunch
rather than dinner we'd need to spark conversation by
raising juicy topics rather than serving cocktails and
wine. The topics of convergence and quality of
service (QoS) did the trick.
The group was split on whether carriers would be
tempted to send voice traffic on the same IP networks
that carry data if the data nets were as reliable as
traditional voice networks.
"The money's coming from voice right now," said
Chatter, whose denim Nexabit shirt bore a striking
resemblance to Dahod's NetCore shirt. He added
that to send moneymaking voice over a data network,
service providers must first be able to guarantee a
certain amount of bandwidth for voice - and that
means shoring up IP networks.
"The jury's still out on whether this converged
network is going to exist," said Bernstein, in between
bites of his gamberetti. "There's still this problem right
square in front of the carriers: They have orders an
arm's length long to build IP networks, and they can't
fill them. Whether they run voice or not, there is a lot
of IP demand."
"No question. I'm not talking about convergence as a
goal," Chatter retorted. "It can be a viable alternative
only if the carriers can make money on it."
On QoS
All the vendors in this game have some kind of story
about QoS, where they can ensure that high-priority
traffic gets through the network faster. The carriers
say they want QoS - but would they even use it?
NetCore's Dahod said the primary focus for carriers
now is just "doing more of what they are doing now at
a lower cost and at a faster rate."
"You've got to go talk to the guys who are running
these networks," said Bernstein, an industry veteran
who has done stints at AT&T and The Santa Cruz
Operation, among other companies. "The guys who
have Cisco 12000s now - they're not using tag
switching. They're hardly using any of the features."
He pointed out that service providers are installing a
lot of these devices to fill their high-bandwidth needs.
"It's quite possible it's the wrong type of intelligence in
the 12000," Dahod fired back. "There's not the right
kind of intelligence for carriers to do the traffic
engineering they need to do to bring business
customers into an Internet infrastructure."
While Chatter agreed that such technology is
necessary, he said carriers need to address bandwidth
constraints first. "If you've got a deadlock in
downtown Boston, you put in a much bigger highway
and you apply traffic engineering to it. But you don't
do it the other way around."
On IP vs. ATM
The vendors around the table have an IP focus, and
they are building routers based on IP. But they fully
realize their devices will have to play well with ATM
gear if they are going to fit into a carrier network.
"ATM switches absolutely will play" in the network
for some time to come, Panditi said. "The customer is
going to decide" how it will migrate from ATM to IP.
"We sometimes get caught up in our own hype,"
Dahod acknowledged. "We all grew up in the IP
world, but nothing takes over overnight."
Grady said the IP community can learn from ATM
product and service vendors, which he said fell into
the trap of developing a solution and then looking for
a problem to fit it.
QoS services on IP may meet a similar fate, he
added.
On sticking around
We saved for last the topic of whether these start-ups
really have any chance of staying independent for
long. After all, companies such as Juniper and Avici
already are funded partly by the likes of 3Com and
Nortel Networks.
"The big guys are sniffing around because they don't
want to miss anything. That's when rumors get
started," Grady said.
But the CEOs said they plan to keep their companies
independent because new and old service providers
have shown their willingness to buy equipment from
small companies.
Contact Senior Editor
Jeff Caruso or News
Director Bob Brown |
