When a former server/storage executive takes over a networking company, you can expect things to change. This process is happening at Extreme Networks, where Mark Canepa began working as CEO in August after a stint as the executive vice president of Sun's data management group. He spoke with Network World Senior Editor Phil Hochmuth about his plans for Extreme, how the company will compete with Cisco, and the convergence of data center technologies. The following is an edited transcript.
What are your priorities as CEO, after a little more than a month on the job?
What I'm going to focus on at least for the first few quarters are operational kinds of things. The company is focused on doing a lot of different things for its size. So the first order of business is really to do some market segmentation. It's a $55 billion Ethernet IP market. We're a $400 million revenue [company]. Let's go find the market segments that make sense and then let's really throw the whole company behind them.
We have the luxury, being our size, to pick some spots. When you're this kind of size, whatever you choose to do, you'd better be the best at it. You have to punch through whatever the big guys are trying to do. That's going to be the game plan over the next couple of quarters. Get the revenue stream headed in the right direction; get that stuff headed in the right direction and then take it from there.
Was Extreme heading in the wrong direction prior to you coming in?
I don’t bring any preconceived notions either from being inside this company, or from the network industry [itself]. I'm not a networking guy per se, I'm a computer guy; I've been plugging things into the network most of my life. I've been building products into the network, so to me, the network is always something that's sitting right next to you. But I don’t really come to this with a lot of preconceived notions. So it's going to be a pretty pragmatic approach to figure out what are our core competencies, and what is the differentiation we are trying to create. And what is the true value of that differentiation? That's the pretty basic question I'm asking [everyone at Extreme].
So the challenge is really that. Let's take the wishes away from this. Let's just be pragmatic and if it makes sense, it makes sense. But let's not delude ourselves about what makes sense.
The CEO you are replacing was one of the company's founders; how do you take over that kind of role from someone, and what has been the reaction by Extreme employees?
The good news is that my past 10 years at Sun, the company was run by the founder. So there are actually a lot of similarities between Scott [McNealy] at Sun and Gordon [Stitt] at Extreme. So from that perspective, I've been in a company that's run by a founder … I learned. And in 10 years at Sun, when it was run by a non-founder, I saw how [it was run so] that's a pretty good perspective [for this job].
It's surprising, when you're in the storage business, how close you are to the networking business. [In storage] we had to deal with applications, with security, with system management. So 90% of the issues are similar. Then there is a big chunk of the storage business that is file-based. In other words, it sits on the file side … That's all IP-based storage, so you are very much into the IP-network problems — latencies and bandwidth and VLANs, and attaching channels, and assurance of service … I would get together with Gordon, and we would start to draw on the board, saying here are the network problems. And as being a user of the network, I understood them all very quickly. So from the technology point of view, aside from all the four-letter acronyms … it's pretty straight forward. Plus I grew up as a technologist. I started out as an engineer. So technology is rarely what is a problem.
What different perspective can you bring to Extreme, coming from a systems/IT company like Sun?
The big piece is that when you are a systems guy working on servers and storage, the network to you is basically a set of wires that move bits. You think of it in very simple terms. But as you start to design more complex data centers, it's obvious that the network has to be another piece of the entire system. And that means it needs to be provisioned; it needs to be managed; it needs to have GUIs; it needs to have CLIs and an interface, just like servers and storage. So it was actually heartening to see that Extreme Networks has taken a big step forward from the beginning in terms of having highly programmable switch capabilities. The XOS operating system is all about open standards-based interfaces that allow you to have [advanced] system management.
What I can bring to the party is kind of that systems mentality. It's to say, geez, it's not the network for the network's sake. The network is there to be part of the larger system that runs applications. Already, it's changed a bit of the character of our discussions. But I'm pretty clear with the troops. I'm not coming in here as a network guy, so don’t expect me to lead you that way. But expect me to put networking in the context of the data center.
If you really look at a network processor in a switch, it's just another computer that's really, really good at moving stuff around. In my previous life, one of the companies I bought was Pirus Networks. They were an intelligent fiber channel network switch, but the issues were the same: you terminated every transaction; you cracked the packets open; you looked at the headers; you made decisions on them; and you sent them on their way. The problem set of an IP switch is not that different from a virtualized storage switch. So that to me felt good over the last several months as Gordon and I were playing around on a whiteboard. And it also made Gordon feel good [that] he knew I was aware of how to deal with these kinds of issues.
What does your background bring to the party regarding Extreme's metro Ethernet business, which has a different set of issues from enterprise data centers?
Even in the metro space, [having come from the storage] market, it's surprising how much you think about metro networks, mainly because most of storage is about disaster recovery. It's about transport mechanism, data replication; Sarbanes Oxley, for example now pushes backup data centers away from sort of the 10-mile range up to the 100-mile range. All of the sudden you're really talking about metro-kind of stuff. And you're pushing the limits of that.
I've had experience with DWDM. With OEMing DWDM technology and talking to customers about lighting up dark fiber and creating that endpoint connection. You're talking about quality of service, if you're trying to get two big storage boxes to do data replication, whether it's synchronous or asynchronous. Nine-tenths of your problems are with the network fabric, and the carrier that sits in between, and their ability to guarantee bandwidth and latency levels. So as a storage person, I understand the uses of bandwidth … or quality of service; I know the burden I wanted to put on to the network. Now it's just coming from the other side, and saying, oh, geez, now I have to deliver that.
How will competing against Cisco, Nortel and others in the network market differ from competing against EMC, IBM and HP in storage?
You have to attack it differently. If you're a Sun, a big integrated systems company, you have lots of different things. Sun's brand strategy was built around Sparc and Solaris. The real primary job was to take care of a piece of a system; you have a big installed base and a big demand creation machine on the server operating system side. And half of your business is simply caring for that.
You had EMC, but they were third in size after IBM and HP. So you have three integrated companies and then EMC. In the network space, there's a lot more pure play. Cisco is Cisco of course. With a company like Extreme, you don't have a whole [group] of other businesses [like Cisco]. What you've got is nimbleness. What a company like Extreme can do that a company like Cisco can't is to move quickly and be nimble . . . . There are some things big companies can do and they do it very well. So when you're in a company this size, you have to realize what big companies can do very well, and just don’t do those things. Then find out what big companies can't do, and then act quickly to do those things. And that's the trick.
Where I learned that . . . was back at HP. We were really tiny compared to IBM [at the time I was there]. And we developed whole mechanism to develop markets where IBM wasn't going to be. If you are a $30 billion company, all of your processes are designed to find markets that are consistent with your size. When I was at HP, we used to say there were whole industries living where IBM couldn't see, because they were just too big to find them. We can measure market segments in a few hundred million. If we spot a few-hundred-million market segment that you can become good enough to get a quarter of, that may not be a big enough market segment for Cisco to go exploit, because they need so many billions of dollars out of the market to make it worth while.
So we have the luxury of finding these places, and we are nimble enough to get there. That's the difference, I believe. Then you have to wrap your messaging around that. You have to aim that at the right customers and win the battle at the customer's office, rather than winning the battle over the airwaves, or through marketing. We can't win that kind of battle against Cisco.
So what are these lucrative, hidden markets that Cisco is not in that you can exploit?
In five weeks on the job, it's a bit early to tell. We're hot and heavy with marketing, tearing apart that $55 billion IP/Ethernet market and finding the places that are consistent with what we do well. So we haven't completed that analysis. Certainly a lot of the places we're in are the places to be. There may be a few places that were in that don’t make any sense. There might be places that we're not in that would make sense. To me, what's most important is that even in the places that we are in, we have to hone in on the sub-sections of those markets that makes sense. We can't just say we are in the Layer 2/3 switch business, which is a $10 billion market, and just go compete there. We have to be crisper. We have to be able to say what is the $1 billion sub-set of that market that we really want to be in. Because the requirements are going to be different. And therefore the things we want to put into hardware and software will be different.
What are this company's core competencies? Do you see Extreme as an end-to-end network provider, or a specialist in certain spots, such as data center networking?
It's a bit of both, but a little bit more of the latter. You want to be able to walk into a customer's office and say, we can help you with your end-to-end needs. But you have to realize to do a lot of that, we have to create a set of partnerships and an ecosystem that compliments what we do. We can help someone build an end-to-end voice network. But how do we do it? We partner with Avaya … For other technologies, [such as security, or small and midsize business], we go look for other partners. So that's going to be the methodology. We're going to pick some spots that we think we can match well with our technology.
In the metro space, we'll be focusing on the aggregation part of the market. I'd rather be a big player at the aggregation space, than a small player at access/aggregation/core. If you think about building up a value chain, you need partners, you need references, and it's tough to do that across the board yourself. And it's the same in the enterprise.
I think where our value proposition best expresses itself is at the core of the enterprise, because that's where you need the most intelligence. That's where you really need insight and control. That's where you need to virtualize. So you pick your fights. You select the verticals and solution sets you want to be involved in.
What is the state of the Avaya/Extreme partnership, given that both companies have new CEOs?
I feel comfortable with the partnership. If you look back at the history of Ethernet in the last 10 years, it was designed as a highly available network. It was designed to survive nuclear bomb attacks — if you think back to ARPANET. It takes apart data and turns them into packets, and sends them over multiple paths, and guarantees they get from one point to another. It makes no other guarantees, other than the last packet will show up eventually. It's hell to send voice that way. You sort of have to have things coming in the right order. And that requires a level of intelligence of a company like Avaya that does nothing but voice.
While Avaya and Extreme are partners, Juniper is also a partner with Extreme. Often these three vendors' components are the basis of some large network deals [such as the Wynn Hotel in Las Vegas].
Does your relationship with Juniper change now that they are an Ethernet switch company?
We're going to have to watch that. The good news in all of this is it continues a healthy state of co-opetition … there are lots of joint implementations where our stuff and theirs is installed.
Is wireless the right business for Extreme to be in? Can it be a leader in wireless LANs?
That's an interesting question. If you look at the wireless market, it's about a $1 billion market. You could equate it a bit to like selling optics. It's a connectivity technology. So we're going to have to explore. There are two ways to think about this. One is the pure technology aspect: do you want to be in the business of developing the hardware, the software the sheet metal?
The other is at the systems level. If you are going to be a provider of sophisticated enterprise access technology, wireless is a very intriguing technology, when you also offer things like access control and security. Wireless is a veritable pool of viral infection. It's a great way for hackers to get into a company; it's all over the place, and you can't control anything. A customer who is nervous about running wireless is going to be asking a lot of questions about security, and Extreme can be the kind of sophisticated company that can come in and help them. So we need to distinguish a wireless [business] from the bits-and-bytes [development], from wireless as a piece of the enterprise [puzzle]. We may end up with different answers. Within the next few months [we'll figure out our strategies as to what we want to be involved in]. So far our strategy has been to partner on a lot of the technologies.
Do you feel advanced switching — high-end, high-speed switching — is still Extreme's core competency? Or has the company gotten away from that?
At first, Extreme's core competency was putting Layer 3 technology into Layer 2-style hardware, and just blowing away everyone with speed. But in practically every industry, you never survive on speeds and feeds. They get you going, but pretty soon you'd better start being in the business of solving customer problems. And most customer problems are not reflected in speeds or feeds. It's fun to watch Intel. It used to be microprocessors were all about megahertz and gigahertz. Now [clock speeds] are gone. Now its called "Duos" "Centrino" … anything but megahertz and gigahertz. So even in the CPU business, which was governed by the numbers of speeds and feeds, is changing.
That doesn't mean we're not going to be right there to ensure that our products are fast. We have people on the standards committees for 100 Gigabit Ethernet, and driving all of that. But you can't run a company purely on speeds and feeds. At one level yes, but at another level we're going to be more of a broader company, focusing on how we can solve customer problems — problems that are application driven. Extreme will be much more focused on what are the applications, and how can we make them run better.
What is Extreme's role in data center virtualization? How does Ethernet switching become part of that?
That is certainly Cisco's rhetoric. When Cisco bought Andiamo, my antennae went up. Then when they bought TopSpin, my antennae really went up. [While I was at Sun], my nightmare scenario was [if] Cisco bought Veritas. I thought for sure they would buy Veritas. Because then all of the sudden, what do you get? You'd have a company with file systems, volume managers and networks that can really suck a lot of value out of the servers. It's the systems vendors' nightmare because then the reality become this: all you need are some cheap Linux blades on one side, and some cheap JBODs [just a bunch of disks] on the other side, and you've pretty much sucked everything out of the data center.
Interestingly, Cisco did not buy Veritas. Maybe they didn't want to become that much of a software company. From the point of view for Extreme, there's a realization that this stuff is going to matter. There is real convergence in the data center. On one level, there is good news there for us because it puts more complexity into the switch, which is something that plays to our core strengths. The second thing is that it puts more different kinds of flows that have to be managed by the switch. So we have an opportunity to stake out a piece, whether it’s the grid that requires this kind of intelligence … or other types of architectures.