FRAMINGHAM — After years of stagnant or lackluster growth, Extreme Networks tapped Oscar Rodriguez to bring new life to the switch maker in 2010.
Can you start by giving us a review of 2011 and what kind of progress you made, and how you plan to map that into 2012?
I came on board at the end of August 2010 … By December 2010 we had a strategy together of what we were going to do and we announced that strategy in January: Extreme was going to focus on specific market verticals. We selected three market verticals: cloud services and cloud computing; education; (and) large-scale mobility. You’ve got to have products that are purpose-built for mobility, purpose-built for the cloud, and the operating system is really now purpose-built for a lot of the things going on in education and that can then dovetail in to healthcare and other things like that. We also selected those verticals to have a halo effect on other things that we were doing in the business so we weren’t going to give up the ghost on the existing business.
What we’ve done is selected places that we’re going to compete from a focus perspective, and what that enables us to do is really two things: focus your R&D on building very specific products for those areas as opposed to trying to address the wide requirements and the wide marketplace; the other is it lets you focus your marketing and market awareness generation in those specific places where you want to win. So that’s really our goal — it’s to really focus in those spaces and not ignore the rest of it … We’re in our new cost structure and have our new strategy going out the door and in to January.
How do you plan to build on what you accomplished in 2011?
In the process of making all these changes as a company, we actually then set our criteria for the products we’re going to build, and we’ve got brand new products that are coming that will be available in the early spring. And as these products come out, they’re very focused on very specific areas. So we’re getting these products out the door, but 2013 is really where we’re expecting to see growth.
What’s at the top of your priority list for 2012?
The first set of goals is to get our products out the door. They’re due in the early spring and I believe that they will be on time with no issues. The second set of goals is to make sure that all of our salespeople are trained to be experts. One of the things that we’re also doing is revamping our sales force, not only retraining the internal folks but adding people from the outside. I’m looking for really great salespeople that want to have a lot of upside. We can bring people on board that have a Rolodex and we can train them to be the black belts in data centers. We’re shaping ourselves to be that kind of vendor.
A lot of the other big players are starting to integrate their networking in the computing side of the house. How do you respond to that market dynamic?
It’s a dynamic that’s bound to happen because if you’re a large comprehensive vendor, you want to be able to offer service capability. In the case of any of these large vendors, they’re really selling a service. You’d have to be able to bring together multiple pieces, so you’re not just selling consulting, you’re selling also product and the know-how of the overall system.
That does not mean, however, that there isn’t a very large part of the marketplace who will say, “I want to be able to choose those pieces for myself, I want to choose best of breed.” Unless you’ve got a very esoteric technology, they would rather select the best of breed for themselves because that way they control their cost of goods, they control their facilities usage, they control their people and they have enough technical depth to be able to run it themselves.
There are a lot of corroborating activities going on in the marketplace that say if you choose from a big brand, you’re going to suffer in one or two areas. Performance of the overall system may be very good, but the individual components may be mediocre. Unless the big brands can overcome the performance perception that they have, and the very real overall economic issues of having sub-optimized systems as opposed to optimized systems, then I think there’s always going to be room in the marketplace for those customers that choose best of breed.
What do you see as your major challenges in 2012 as you build upon what you accomplished in 2011?
Customer perception – making sure that we drive our customer perception as quickly as we possibly can. The more that we build customer awareness of our products, the sooner you get into the game, the better off you are. The other thing is having the denser boxes. You lower the power consumption per port, you make the network physically smaller so you’re using less racks to run the same size of network. With our BlackDiamond X8 you can fit nearly 2,300 10G ports in a rack. That’s unheard of — nobody can do that.
So if you’re a cloud guy, you want the network to be as small as possible, use up the least amount of room, because that way you have more racks for servers, more racks for virtual machines, more racks for storage, and that, in essence is what you sell — you sell virtual machines and you sell storage. That maximizes your revenue out of the same facility size, and lowers your overhead. So we have to make it as small as possible and as quiet as possible so that these cloud guys can pay attention to the important things, which is driving more revenue and driving more virtual machine space.
Is some of that to say that you’re advocating for flatter networks like a lot of the other guys do as well?
It’s got to be flat, you’ve got to reduce the numbers of tiers. The old multi-tier way of doing this doesn’t work. You’ve got to be able to have active-active switching. You’ve got to be able to have multi-path capability. What it does is it lowers the number of tiers, makes things highly redundant and fault tolerant, and if you can then put it in a nice, dense package, you’ve got an advantage for a cloud operator that’s unheard of.
Juniper is out there talking about the need for single-layered – not even two-tier stuff, but do you go that far? Or do you think two tiers will suffice for many customers?
Yeah, the BlackDiamond can do one tier, and because it’s so dense, we can do one tier with a single box. The way Juniper has done things, they’ve said, “OK, do all these multiple boxes, we can make it look like one tier.” The problem is, now you’re locked in to Juniper. So I kind of look at Juniper like the way Bose has done their audio system — very high fidelity, no doubt about it, but you’ve got to cable your house differently than everybody else does. You don’t want to do that non-standard cabling because now you’re stuck. You’re always going to buy a Bose system no matter what.
How many server nodes can you support from a single switch?
We can do 768 10G server ports on one (chassis) so that’s three times that in a rack. Or 192 40G ports. We were first to market with 40G. We now have the most comprehensive portfolio in 40G capability, and we’ll continue to push on that and continue to bring more 40G capability out.
If you look at the combination of 10G and 40G, that’s really being driven by what’s happening at the server level… thanks to Intel’s work on moving 10G onto the motherboard. So any high-performance data center and high-performance cloud capability is going to be 10G at the server with very speedy processors to maximize the number of virtual machines. And then that network is going to have to provide 10G access to the server and then 40G trunking so that you can operate in a fully meshed environment and be able to give any server access to any storage or any other server at any point in time. When virtual machines move around you never can tell which physical server they’re going to sit on.
As you focus on data center cloud fabrics going into 2012, are there any pieces that are missing right now that you need to have in order to fulfill that mission?
One thing I think is very important is policy management and networks. We’ve had the luxury in the networking industry for many years to just throw more bandwidth at the problem. As soon as anything got congested, just throw more bandwidth. As soon as you needed to have mission-critical information, add another switch. If you needed to have quality-of-service issues, you could just give it tons of bandwidth and it will resolve itself.
It’s getting to the point where, thanks to applications that are coming from the mobility side, every iPad and every Android (device) is going to have a different set of applications per user and so the traffic is no longer predictable. You’ve got to have policy that’s applied that is based on the criteria of what we should be allowed to do on the network, the types of resources we should be allowed to access, the kind of bandwidth and quality of service that we need to have, and that needs to go all the way back into the data center.
We’re going to see quite a bit of policy management start to be required and necessary, in addition to security management inside the data center for intra-data center productivity between applications and VMs, and the network has a big part to play there.
(From Network World U.S.)