How Gigamon’s Founders Bootstrapped a Networking Hardware Company to Profitability
(Page 2 of 3)
tap into their network switches. The problem is that even the most expensive switches from Cisco and Juniper come with only one or two external ports each for this purpose, and the switch makers do very little to help customers tap or replicate data flowing through their devices.
“If you replicate traffic, that will usually cause a performance degradation of 1 percent to 5 percent,” explains Ho. “That may not sound like much, but when you’re competing [with other switch vendors], even 2 percent is a lot. It’s a problem all switch vendors have, and they understand it, but they don’t want to address it, because from their business point of view they don’t need to.”
That’s where Gigamon comes in. The company is like the guy who brings a power strip to a technology conference where there’s a shortage of wall outlets: he’s suddenly very popular. Gigamon’s gadgets plug into Cisco and Juniper’s switches, where they dip into the data stream without slowing it down. On the business end of a Gigamon switch, there are a bunch of extra plugs where administrators can attach other network monitoring equipment.
But “we don’t just get the data and feed it to you,” Ho emphasizes. “We really pick and choose the data you want. For example, if you have an appliance that’s only for voice traffic, we can give you just voice. Another guy might say ‘I’m only interested in HTTP, so we get you that data.” The technique for diverting specific chunks of data based on their content is called deep packet inspection, and it “gets you from a firehose down to a garden hose,” Ho explains.
It’s a pretty useful trick, and it makes Gigamon the go-to provider for almost any big organization that’s concerned about monitoring its network traffic. But it took the company a couple of years to figure out how to build its switches—which have some heavy-duty processing power built in—without spending a lot of money. For the most part, the company steered away from custom chips or ASICs (“application-specific integrated circuits”) in favor of cheaper, more flexible processors based on field-programmable gate arrays, or FPGAs. “By using these off-the-shelf chips, we could deliver not necessarily 100 percent of the functionality, but if we can deliver 95 percent, why not?” says Ho. “Time to market was the important thing.” It also settled on a basic chip architecture that’s easy to scale up without requiring a big redesign. “We went from 1 gig (gigabit per second) to 10 gig without redesigning anything,” says Ho. “It’s cheaper in terms of development, because I can have a new generation in a matter of six to eight months instead of two years.”
Gigamon introduced its first product in May 2005 and sold just 60 switches in its first year, but managed to hit profitability starting in 2006. (Along the way, Ho says, the company filed for eight patents on covering packet replication by ASIC and FPGA chips.) Today Gigamon has several hundred customers, including all of the major Internet service providers and … Next Page »