Juniper's entry into enterprise switching with the EX line is rooted in extending a common operating system across the switching, routing and security domains of an enterprise network -- something that's lacking in what's viewed as a mature market dominated by Cisco. Juniper CEO Scott Kriens shared his thoughts on the company's opportunity -- and what it means for Cisco's current competitors -- with Network World President and CEO John Gallant and Managing Editor Jim Duffy at this week's EX launch in New York.
Why develop the switch on your own instead of getting it through acquisition?
The power of [our Juniper Network Operating System (JUNOS)]. Wanting to see a common platform and one feature written one time running everywhere on the same day. From switches to routing and even security products now. The development decision was driven by the JUNOS operating system strategy. In the case of operating systems, more is worse.
The switching market is considered mature with little opportunity to penetrate Cisco's dominance. Why do you feel differently?
For the same reason we felt that way 10 years ago in the router market. The requirement for the network has changed. What used to be networks simply for the sake of connectivity are now networks for the sake of survival. In that scenario, the performance of the network matters more, and the criteria for buying and operating a network changes from what it was to what it needs to be. And that, to us, is an inflection point that we're very familiar with. And we're seeing, as Yogi [Berra] says, Deja vu all over again.
So the seven other competitors to Cisco in enterprise switching did not see this? Or could not address this in some way?
Well it's the synthesis of the portfolio really. Because it's switching seamlessly, interoperating with routing, with security and so on, that I think is the distinction. And this is the difference between the box strategy and an operating system strategy, and it's that latter strategy that is unique relative to all the players in the market, big and small. That's what our customers have told us is the distinction and so that's what we're investing in. The best way to think of those products is as vehicles for that operating systems strategy.
Has there been a fundamental shift in switching that makes now the optimal time to enter the market?
I'd almost call it a different market. Because it's not a switch market; it's a network services market, which is why the network operating system is so important. That's a very different model than thinking of it as a switch or as a hardware problem. Different architecture, different priority of development. That's why it has to be so seamless through networks where you're connecting the switching, the routing -- it's all about the operating system.
Is that why it's not a concern that the switch capacity of the EX 8216 falls short of Cisco's Nexus 7000 and only matches Foundry's BigIron RX?
The important metric to look at is actually the packets per second that can be processed inside the machine. So you can physically connect ports and add up the total speed of the physical ports, but once you get into the machine it's not the port hardware speed that matters -- it's how many packets per second you can process inside the machine. That's really where you hit the capacity limitation. So that's where we think we have very strong performance capabilities. We have over 500 patents. And most of them are centered around that problem. That's where I think the distinctions are going to be drawn.
Why did you not aim for a Nexus-level system in terms of its 15Tbps switching capacity?
That's the aggregate speed of the cables. If you look inside that machine I think we process twice the number of packets per second. That's actually what matters to anyone using one of these machines as opposed to specing it. It's not how much traffic can I plug into it on the outside, it's how much can I get through it. If I plug in 10 times more than I can process then I don't use that -- I can only use what the machine can process. At roughly 120 million packets per second of processing with our machines, that's twice what you can process with machines that -- I don't know if they exist or they're just talking about them or what, but -- even if they do, that's half the packet processing capacity that we have.