When Cisco senior engineer Tony Li lost his temper and nailed his resignation to his boss' door in April 1995, his act may have been loaded with more symbolism than even he knew. Almost a year later, the high-end network router Li had been working on is still not available and Cisco's competition is starting to gain ground. The $5-billion-a-year networking colossus suddenly finds itself in an unfamiliar position: behind.
Signs that Cisco has lost its technological edge have become increasingly hard to overlook. Mid-January saw PSINet, one of the nation's largest Internet service providers, unexpectedly choose routers from Ascend over those from Cisco. Then, in a Feb. 11 report, Wall Street analyst Gina Sockolow revealed that Cisco has been pleading with customers to delay purchases from other vendors until Cisco can get its own offerings out the door. And on Feb. 17, an industry trade journal published a leaked Cisco memo that suggested the company was stalling an Internet standards committee in order to buy time.
The story of how this long-dominant company ended up in such a humbling position reads like a morality tale for young engineers. Less than two years ago, Cisco stopped focusing on simply building better routers and turned its attention to a new marketing strategy. The goal was to lock in customers by using proprietary protocols so that Cisco equipment would work better with other Cisco products than with its rivals'.
But today those protocols are still just brochure-ware, and Cisco's routers have become increasingly dated. And, in the ironic twist that accompanies every good moral fable, Cisco's marketing strategy is now actually hampering the company's growth.
Routers are specialized computers that direct data through the Internet. High-end routers are like race cars: incredibly expensive, finicky and rare. And that's where all the interesting challenges are first tackled and where the top engineers gravitate. Cisco's current top-of-the-line router, designed in part by Li, is the 7500. Fully loaded the machine goes for about $100,000 and can route a million packets of data to their destinations every second.
For a while, that capacity set the industry standard. But within months of the 7500's release in 1996, some Cisco customers were describing the machine as "long in the tooth" and "dinky." With the Internet continuing its steep exponential growth, backbone operators like MCI and UUNet knew they would need routers that could handle 30 million packets per second by the end of 1997.
Cisco's engineering group started drawing up plans for their next-generation router. Called the "BFR" (which stands for "Big Fast Router" in the PG-13 version of this tale), it was seen by Li and other top Cisco engineers as an opportunity to build something that wouldn't be instantly obsolete. Forget about 30 million pps! Using clever techniques that had so far only been discussed in academic journals, they could far surpass that target.
But Cisco wasn't interested. According to what Li told his friends after leaving the company, Cisco chose a very conservative and evolutionary architectural approach--one that both failed to meet the needs of its customers and opened a window of opportunity for its competitors.
Li refused to comment for this article, but what makes this account sound like more than sour grapes is that it's consistent with Cisco's standard operating procedure. Whereas companies like Intel and Microsoft have large research labs that are charged with looking three to five years into the future, Cisco has always kept its engineers focused on the nearer term. Then, if some clever start-up suddenly leapfrogged Cisco's technology, Cisco would buy the company.
But lately Cisco's competition has been snatching up such companies first, as Ascend did in buying a small company called NetStar last year.
Moreover, it is disgruntled former employees of Cisco itself who are producing the new technology elsewhere. Li and fellow Cisco alum Paul Traina are now at Juniper Networks, a year-old networking start-up with $8 million in venture capital. Not much has leaked out about Juniper's forthcoming router, but the few details that are known have ISP operators salivating.
Another Cisco error, evident from a reading of the company's Securities and Exchange Commission filings, was its decision to emphasize its proprietary software.
The company seems to have understood that it would be unrealistic to plan on having the best networking product in every category, ranging from switches to remote-access devices. There are just too many. Instead, it is hoping to leverage its strength in a few product categories by having all its products support the same proprietary software. That software is the vaguely defined but much-hyped CiscoFusion. When an ISP manager knows that his existing Cisco router works best with a Cisco switch because they both use the software, the theory goes, Cisco's switch will be preferred over the competition even if the competition is technically superior.
It's a strategy that worked well for Microsoft. But networks and planned incompatibility are inherently at odds. If the history of networking says anything, it's that open standards win over proprietary ones every time. And while it's not at all clear what the benefits of CiscoFusion will be for customers, its immediate disadvantages for Cisco are becoming apparent.
Take Granite Inc., a company that was bought by Cisco in September for $220 million. At the time, Granite was well on its way to developing a gigabit ethernet switch--widely seen as the next big thing in local area networking. But today you'll find Granite engineers stuck in Cisco's integration lab as they struggle to make sure that their hardware can talk the esoteric protocols that make up CiscoFusion. Whereas other vendors who only had to worry about well-known and open protocols will announce their gigabit ethernet products this month, Granite doesn't plan to release its until early 1998.
That demonstrates that even if Cisco decided to buy a company such as Pluris--a one-man start-up with a revolutionary new router architecture--it would still take the company too long to bring it to market. This is the parable's kicker: Cisco's software-based business strategy is incompatible with its research-through-acquisition strategy.
How this morality tale will turn out is hard to predict. Cisco has dug itself into a deep hole, but its enviable distribution channel and massive installed base may buy it the time it needs to get out. The question is just how long an ever-more-demanding customer base will accept seemingly clever business strategies in lieu of true innovation.