A mere decade ago, IBM was king of the corporate networks. That was no surprise, given that IBM essentially invented the mainstream corporate network. The surprise was that before 1999 was out, IBM was out, too -- out of the network business. The company had been dethroned and exiled, finally selling what was left of its network business to Cisco.
As is so often the case, it was IBM's massive success in corporate networks that led to its downfall. It simply made too much money selling traditional network gear that had been developed to implement IBM's breakthrough Systems Network Architecture (SNA). Although that architecture was introduced in 1974, it succeeded in becoming the common network architecture for virtually all of IBM's customers.
At the heart of the SNA network was the beast known as the FEP -- accurately describing its job as the Front End Processor. It was the size of a refrigerator, and when I first saw one in 1980, I noted that it was the last box in the computer room that still looked the way a computer was supposed to look -- lots of flashing lights. It was quite complex (I know this, too, from firsthand experience) but it had two qualities that IBM probably cherished most: It was an indispensable part of the network, and it was expensive. The FEP was a cash cow.
Unfortunately the brave new world of the Internet was going to do away in the end with the FEP -- and all that revenue. And although there certainly was money to be made in selling internetworking gear (as so many vendors have since proved), IBM apparently didn't want to give up the ghost on FEP money until every bit of it had been wrung out of customers. From my observations at the time -- and insights from those involved -- IBM's internetworking efforts were hampered by the power of the pro-FEP group. In the end, IBM chose not to cannibalize its FEP sales. Its progress in the Internet router space was limited to a too-little-to-late offering: the IBM 6611, which basically was a product of Proteon (remember it?). So, IBM didn't eat its young, but Cisco and others did. The rest is history.
After my recent travels around Silicon Valley, I couldn't help but think that Cisco is or will soon be faced with a similar situation. Speaking with executives at various companies known for core switching, access routing and security devices, I heard the same refrain: "Cisco is vulnerable because their main goal is to sell more ports and more line cards to trigger customers to buy more Catalyst 6500 boxes -- even though that is not in the best interest of their customers."
Specifically, the executives point to two areas: First, the availability of reasonably-priced 10Gbps Ethernet ports lets users get a lot more aggregation bandwidth with fewer ports. So, if you can get a single 10Gbps port for less than the price of ten 1Gbps ports, you are saving money and slot space. Good for users, maybe not so good for Cisco. These vendors note that Cisco seems to recommend hardware that requires lots of individual ports.
Second, vendors note that advances in the power and capabilities of off-the-shelf Intel platforms are going to reshape how we build our networks and reduce the cost dramatically. High-performance, enterprise-class security devices and access routers, they say, will be able to outperform the ASIC-based, pricey "service modules" that customers now buy for their Catalyst chassis.
So, will Cisco be able to see the world in a way that does not revolve around its current Catalyst? If it doesn't, others certainly will.