Acquiring Growth Networks will help Cisco build a real Terabit router, but would Power X have been a better buy?

February 18, 2000

4 Min Read
Is Cisco Buying the Wrong Startup?

When Cisco Systems Inc. . http://www.cisco.com announced plans to buy Growth Networks, Inc. http://www.growthnetworks.com on Wednesday, it looked as though it had pulled off quite a coup.

For $355 million in stock - small change for Cisco - it bought some next generation switching fabric that promises to boost the performance of its top-of-the-range routers by an order of magnitude (see Cisco Buys Growth, Literally).

But will the acquisition of Growth Networks enable Cisco to catch up with the competition, in the form of Juniper Networks Inc. (http://www.juniper.net) and a bunch of Terabit router startups?

Or is Cisco buying the wrong company?

One of Growth Networks' closest competitors, Power X Ltd. http://www.px.uk.com, a Manchester, UK, based startup, says it is and that Cisco will end up regretting its choice. It says the architecture used by Growth Networks has an intrinsic problem: it can't guarantee low latencies.

And that's not sour grapes, according to Russell Johnson, vice president of sales and marketing at Power X. Johnson says that Cisco has done everybody except itself a good turn by purchasing Growth Networks.

"Customers have been holding back, waiting to see which startup was going to be taken out of the market by Cisco. Now they know, it'll free up the market. And with Growth Networks gone, we're the only show on the road that's still totally independent," he says.

Other companies developing next generation switching silicon are also OEM vendors, Johnson says. As a result, a lot of vendors would prefer to buy from Power X rather than go to a potential competitor.

The intrinsic problem in Growth Network's architecture has already been spotted by some vendors, according to Johnson. "A lot of the companies that started talking to Growth are now knocking on my door," he says.

To put things into perspective, Growth Networks, Power X and a bunch of other companies including IBM are developing next generation switching fabrics for use in Terabit equipment at the edge and in the middle of carrier's optical backbones.

The gear at the edge needs to combine packet, cell and circuit switching so that it can act as on-ramp for existing services, such as TDM voice circuits, ATM and IP. (These are the multi-service provisioning platforms surveyed in Sonet Goes POP). It also needs to handle larger numbers of channels running at OC-48 (2.5 Gbit/s) today and OC-192 (10 Gbit/s) in the future.

The gear in the middle of the network will switch big pipes. Sooner or later, it's likely to be replaced by pure optical devices, ones that use microscopic mirrors or other technologies to divert wavelengths onto different routes without converting it back into electrical signals (see Optical Illusions).

Growth Networks is developing silicon for switches in both markets, using multi-stage crossbar technology. This enables each chip to support more than a thousand ports. But the architecture suffers a big disadvantage: It can't guarantee low, fixed latency because traffic has to contend for access to the switch itself, according to Power X.

Power X admits that its single stage fabric architecture supports less ports - 256 - but says its technology is deterministic. There's no contention for access. This guarantees latencies measured in tens of nanaseconds, something Growth Networks can't deliver, according to Johnson.

Power X's chip is also much smaller than Growth's, and consumes far less power - important considerations for equipment developers, according to Johnson. (For more on how Power X compares itself to Growth Networks, see http://www.eet.com/story/OEG20000124S0043)

Johnson's comparisons are "a marketing ploy by Power X to garner some much needed attention," according to Steve Mullaney, Growth Networks' vice president of marketing. "What do you think are the chances that Cisco didn't think about latency issues when they evaluated our architecture?" he asks.

From Cisco's point of view, it already has a switch for routing wavelengths across optical backbones, thanks to its acquisition Monterey Networks for $500 million last summer.

The acquisition of Growth Networks aims to boost the performance of its top-of-the-range GSR 12000 routers, which handle IP traffic, not wavelengths. "It will allow us to scale our architecture to tens of Terabits," says John Doyle, chief science officer for Cisco's service provider line of business in Europe, the Middle East and Africa.

Doyle points out that Cisco has made two previous acquisitions aimed at enhancing the GSR12000's performance in carrier backbones. Its purchase of Skystone Systems Corp. for $66.5 million in 1997 gave it an OC-48 Packet-over-Sonet (POS) interface. Its purchase of StratumOne Communications for $435 million last year gave it the OC-192 POS interface that was demonstrated last December by Qwest Communications http://www.qwest.com

. by Peter Heywood, international editor, Light Reading http://www.lightreading.com

Subscribe and receive the latest news from the industry.
Join 62,000+ members. Yes it's completely free.

You May Also Like