Riverbed CEO Jerry Kennelly talks WAN optimization

In an interview, Kennelly discusses the future of Riverbed and how WAN optimization is like building a private cloud

If you think WAN optimization is a niche market, don't bring it up around Jerry Kennelly. Co-founder, chairman, and CEO of San Francisco-based Riverbed Technology, Inc., Kennelly is a fervent believer that WAN optimization is the foundation for the next generation of IT infrastructure and that Riverbed is poised for a dominant role not only in corporate data centers but in the cloud as well.

Since its founding in 2002, Riverbed has become the leader in WAN optimization (according to consultancy Gartner Group, Inc.), and it continues to grow at a rapid clip, as in 39 percent year-over-year revenue expansion in its first fiscal quarter ended June 30. In this installment of the IDG Enterprise CEO Interview Series, Kennelly spoke with IDGE Chief Content Officer John Gallant and Network World Senior Editor Tim Greene about battling with Cisco, the expansion of Riverbed's product line, and big opportunities in the cloud.

IDGE: Where does Riverbed go beyond WAN optimization? You are a dominant player, but the danger is that you become a one-trick pony. How do you expand the scope of this business?

Jerry Kennelly: What we're really doing is layer-seven application acceleration, and that has much deeper implications than simply making a particular land line faster and cheaper than it was.It's something that changes the nature of global IT infrastructure for every major company in the world. Everyone likes a fast line. It was attractive to people because it saves them bandwidth. It's much cheaper to do optimization and compression across the network than to buy bigger links. But then we saw people doing data center consolidation with it, which is moving all the server and IT infrastructure out of branch offices, out of multiple data centers into just one or two. That trend has driven a lot of our growth in the last three years. Our products make that possible because you can't do data center consolidation unless you can give reasonable performance to the people who no longer have local servers.

We woke up one morning about six months ago to discover -- wait a minute -- what we're doing is creating private clouds, because what data center consolidation does is the creation of a private cloud. So, in fact, we've actually penetrated the cloud market.

The further implication is that if you have to have our technology to do private clouds, well, guess what, you can't do public clouds without it either. The biggest companies in the world -- the biggest service providers, the biggest systems integrators, the biggest Fortune 100 companies -- are coming to Riverbed to ask us, "How do we do our cloud infrastructure into the future?"

IDGE:Cisco started out in the multiprotocol router business and expanded into many other areas, including switching, storage, security, and more. Should we expect to see a similar, diverse growth track and expansion plan for Riverbed?

JK: Cisco is the layer two and three of networking, but the action now is less at layer two and three and much more layers four through seven.  Cisco's the king of two and three and we're the king of four through seven.

We just have a huge future for ourselves. We bring the capability customers could only dream about, that no one thought was ever possible, and here we are delivering it. There is a big product future, revenue future, for us. We're a strategic partner. I talk to CIOs all the time now, and the importance of having knowledge workers connect to their applications effectively, cheaply, globally, seamlessly, 24-7 is critical for them. That's what we do. That gives us an incredible position going into the next decade.

IDGE:Where could the company go in that layer four through seven realm?

JK: Our big concentration now is the edge of computing, the branch offices where a lot of the workers are. But there's an interesting thing happening with data center consolidation and cloud computing. The old paradigm was the corporate headquarters was where the corporate data center was, and all the branch offices were the edge. What's happening now is the corporate data center is being moved to a cloud somewhere else. It's in Wyoming or Arizona. What that means is that the corporate headquarters is now part of the edge. Having the headquarters join the edge increases our market opportunity.

We introduced a product about a year ago called Riverbed Services Platform (a virtual server environment on Riverbed's Steelhead appliances) that allows you to plug in five other software modules of services you would place in your branch office. We're approaching the ability to give people that dream of the one-box in the branch office. You don't need a separate box to run your Windows server anymore, or really any other server in your branch office.  That becomes the beachhead of a very big market. Into that single box you can plug in your security, your media server, your media streaming, your DNS, DHCP, domain controller, your print server, your Microsoft Windows server, anything you want. That becomes a large ancillary and synergistic market. That's the edge.

Then we have a market in the data center, the connection back to the edge. You have to have a Steelhead at either end of the link. So we have great big honkin' Steelheads -- one of which we just introduced, the 7050 -- that allow for massive scale in the data center for people to connect to the branch offices. Then further, the connection to the backup and recovery data center for storage to be sent over the WAN instead of by tape. Put it all together. We have the edge market, we have the data center market, we have the connection between them, and then we have the backup and recovery market. All available to Riverbed, just from that little start up making stuff go faster across the wide area network. The implications of speed and performance turned out to be much vaster than anyone dreamed back when we appeared on the scene in 2002.

IDGE:Playing that Cisco analogy a little farther, it grew primarily through acquisition. Is that the model for Riverbed?

JK: We've grown organically through our inside development and we desire to do acquisitions. We acquired Mazu Networks, (whose products) we now market as our Cascade product. That's a network performance management and security product that plugs in very naturally with the Steelhead. We seek to do more acquisition. It's tougher. The whole layer-seven world is still somewhat new so there are fewer attractive targets for us in terms of acquisition. But we're very interested in doing that, and we look very hard. We have a high bar about what we would acquire, obviously, having just done one in nine years. But we've got plenty of cash, we've got plenty of management bandwidth. If we can find good targets, we will expand by acquisition as well.

IDGE:Speaking of risk, other technologies in the past have become simply features that live on other devices. How long before that happens with WAN optimization?

JK: Cisco, our biggest competitor, hasn't done very well in this market.  They were going to put [WAN optimization] as a blade in a router.  Well that's not a natural place for it.

Basically we put a powerful server at either end of a wide area network link that has a lot of processor, disk, and memory. And you need all those to do the magic of compression, de-duplication and acceleration. Routers and VPNs, basically they're just little CPUs, hardly any memory, no disk, hardly any compute processor at all. In fact, the box that supports a router has nothing whatsoever in common with the box that supports network acceleration. It's much more natural to stick a router in [acceleration gear], and we've done that.  We announced it within the last three months, a partnership with Vyatta, which makes a small router that just plugs into a Steelhead box. It's more natural to put that in a slot in a Steelhead box than it is try to put a big server tacked kind of artificially onto a router. Cisco has this ISR box where they have a slot for a WAN optimization board. The only thing that's integrated in that scheme is the power and the chassis. There's no other integration.  It's a completely separate computer that just happens to share the power supply and be wrapped in the same piece of sheet metal. If you made it big enough you could park a Toyota in there and have it be a router, a WAN optimizer and a Toyota garage.

With our approach, where we've got a VMware layer on a powerful, it's a more natural fit to put a router and a VPN and a Windows server and whatever else you want to do on top of the Steelhead than it is the other way around.

IDGE:You're doing well making money off data center consolidation, but how much longer do you think that will be a cash cow for you?

JK: It's two a.m. in a 24-hour day. Most of the estimates for the market are six to eight million branch offices worldwide that are connected by network links. Everyone in the [WAN acceleration] market together has about 250,000 or 300,000 of those in total, so there's a vast amount of market left to come. It's still a young market. Riverbed is nine years old, but if you look at our life-to-date revenue, which is something like a billion and a half, a billion of that all was achieved in just the last three years, we've been on such steep ramp. The functionality and the economics are so attractive that we think the adoption rate ultimately will be much higher. We're not claiming a hundred percent, but it's reasonable to think of 50 to 60 percent adoption rate for this over the next 10 years. That throws off a tremendous amount of revenue.

Again, these data center consolidations are a proxy for private cloud computing and then public cloud computing. The genie's out of the bottle on that. It's not going back. The ability to connect at the application layer across networks is going to be a permanent requirement of everyone. We've benefitted by the fact that latency never goes away. Einstein was right; you can't exceed the speed of light. As long as people are geographically distant from their compute resources, they'll have to have our technology.

IDGE: Recently Riverbed announced the Virtual Steelhead, and you also have the RSP. They seem to compete. Can you distinguish the different use cases? 

JK: In general we find IT guys want to buy appliances because it's a lot easier for them. We found there are certain corner cases in the market where people want specialized server form-factors because of space requirements, heat. The military has special ruggedized server boxes that can take a bullet and be dropped out of a helicopter and operate in dust and a 115-degree heat. We don't make that box, but they want to have a Steelhead out there anyway. So the Virtual Steelhead allows that. There are places in boats and submarines and oil platforms and news vans and trailers and construction trailers, again where people have a limited space and they like the Steelhead functionality, but there's really no room for a Steelhead box. The Virtual Steelhead has to address that market, which is really a corner case of the places we can't fill with our appliance Steelheads.

But there's a second market. Virtual data centers where people are going more cloud, have made dense server farms for their data center, don't really want a foreign appliance in there and just want to run the Steelhead that will connect them out to their branch offices, mounted on their dense blade server infrastructure, their data center in a virtual way. So the virtual Steelhead is for those two markets.

IDGE: You talked data consolidation as being a proxy for private cloud, which we'd push back a little on. Having a consolidated data center doesn't necessarily mean that data center has the characteristics of a dynamic IT infrastructure. 

JK: Either way, we get their business, whether they do it in the cloud format or in a very traditional corporate data center format. But yeah, I take your point. I'm not trying to push the cloud, by the way.

IDGE: Let's talk about that private cloud, though. What's the next phase of growth for cloud and how does that affect your market?

JK: What people want is less and less at the edge because it's expensive to deploy it or you have multiple instances, it's less secure, you have to patch it, you have to back it up, you have to have system admin personnel and resources. It's difficult to keep trained people over long periods of time. Everything that can be pulled back from a branch office and kept at that central data center is what people want to do. People will have a single access point to the network from their data center, then backhaul internally off their branch offices to reduce the security concerns. You'll see more online data available to be backed up across networks instead of through tape backups. So the ability to do disaster recovery, backup and archiving, never touching a tape, from big data center to backup data center across the WAN link will become more important, and that'll be an important market opportunity for us. Then, the density of these big data centers gives us an opportunity to put our biggest machines to work. Data center consolidation will continue, whether it's the old fashioned way or the new way.  It's just what everybody wants to do. It's where the big ROI is. 

IDGE: Riverbed's talked a very little about something called Cloud Steelhead. What is that?

1 2 Page 1
Page 1 of 2