Juniper Networks has always been about high performance and, since it straddles the carrier and enterprise markets, has an interesting perspective on where these worlds intersect. Network World Editor in Chief John Dix caught up with Kim Perdikou, EVP and GM of the Infrastructure Products Group, and David Yen, EVP and GM of the Fabric and Switching Technologies Business Group, to discuss converging needs, tech trends and the company's Stratus project, a single-layer network architecture.
What kind of overlap do you see with carrier and enterprise networks today?
Perdikou: There's a lot of discussion around cloud computing, with service providers preparing to be cloud ready and enterprises [trying to figure out] what they can leverage from a service point of view. In addition, because of the cost and the complexity of running enterprise wide-area networks, many companies are starting to outsource that to service providers. But for service providers, the network costs are rising faster than the revenue they make on them.
Why is that?
Much of the service providers' business model for bandwidth is all you can eat for a certain price. And that is in danger of breaking. We believe it will break over the next three to four years. How do they transform their business model in order to charge for the right services at the right time that people are willing to pay for it? Let me give you an example. I want to pay a certain amount for bandwidth to my home, but if I'm in the airport about to get on a plane, if I could press a magic button and have all my e-mail downloaded in a minute, I would be willing to pay five or 10 dollars for that. I don't put that in the same context as, "How much am I paying for bandwidth for a month?" So how do you build a network that you can start to charge where it's relevant, that people are willing to pay for?
Is part of the answer to migrate everything to IP?
The vision of one IP infrastructure for all traffic has been accepted as how we get there. The problem is legacy services or at least legacy access. It's all about reducing the time from concept of a service to marketing it and making money out of it. And to do that, they really need to redo the network to make it simpler, to make it cheaper to run. Even on the billing side. One customer I spoke to was celebrating that they went from 300 to 30 billing systems. So, if you have that kind of complexity, getting to one network infrastructure and one billing system is a huge, huge challenge.But there is agreement that IP on Ethernet is the Nirvana vision?
IP Ethernet is the long-term answer, but it's not the only technology. In some cases they'll just use optical transport and put routing or switching around it.
So are enterprise and carrier networks getting more similar?
We focus on customers with high-performance networking needs. If you look at the enterprise switching market, it's probably $21 billion, and we believe about half of that is addressable as high performance. Why? Because the requirements of the enterprise are moving more to what service providers require -- scale, performance at scale, availability. I said to a financial company, "You know to a service provider a 911 call is life and death" and he said to me, "a financial transaction is life or death." So, when the customer believes that, the requirements are coming together.
Yen: I just came back from Greater China Mainland and Taiwan, and while it is not going to change overnight, the focus is on growing IP, growing Ethernet.The announcement you made recently about the MX 3D was about scaling Ethernet for carriers, right?
Perdikou: Yes, and it's interesting because when the market for carrier Ethernet started to open up, people that had switching for LAN environments said, "Oh, we can use these existing products there." One, we didn't have switching at that time, but two, we didn't believe that anyway. Ethernet is the transport but you have to build in flexibility and programmability, and that's what we did with the MX3D. We believe it's the first universal edge in the industry that truly meets what the universal edge was envisioned to be. If you're going to deliver a lot of video traffic, you will have fewer subscribers with a lot more bandwidth each. If you're going to deliver mobile voice traffic you want a large number of subscribers that use a small amount of bandwidth.
So, the fungability between services and subscribers and bandwidth enables service providers to build networks that are open and flexible with the right level of security and the right level of functionality. And the fact that we have Junos Software in the MX and across the product lines, and in security and in the switching, is another industry first and differentiates us in total cost of ownership in a huge way.
I mean, just looking at the MX3D, the total cost of ownership is at least 40% percent less than most of our competition. And the MX3D will use a tenth of the power of anybody else that delivers 2.6 terabytes at the edge. That is huge for helping to transform the business model for a service provider and to reduce the total cost of ownership in the enterprise.