Brocade CEO says Ethernet fabrics will undermine Cisco's dominance

Brocade swallowed Foundry Networks a few years ago and has been hard at work ever since marrying the best of Brocade's storage network assets with Foundry's high-end Ethernet network tech. The result, says CEO Michael Klayko, is an emerging product line that positions the company perfectly for the demands of the virtualized, cloud-ready data center, and perfectly to steal share from Cisco. IDG Enterprise Chief Content Officer John Gallant and Network World Editor in Chief John Dix caught up with Klayko at the swank new Brocade campus in San Jose to get the low down.

Read other interviews from this series.

We just toured your data center and saw lots of evidence of Foundry, the company you acquired in 2008. Give us an update on how the Foundry assets are fitting into your strategy.

Brocade was always data center oriented in terms of storage assets, and Foundry had a history of high density, high performance Ethernet. So what we've been able to do is take our storage know-how and apply it to the Ethernet space to improve resiliency and efficiency, and the culmination of that multi-year effort is just starting to roll out in products.

Brocade data center switches take on Cisco, Juniper

I always ask people, tell me what's the most innovative product that's come out in the last decade? Everybody kind of scratches their head and says, "Well, it's been incremental." But there hasn't been tremendous innovation. Storage area networking was fairly innovative, but on the IP side there hasn't been a reason to be. It's been dominated by one player that didn't have to do massive innovation because it had such a large customer base. 

Foundry had a little niche in a high performance, very dense application. Now we're taking data center robustness to that Ethernet product set to create what we're calling Ethernet fabrics. We believe this is the most innovative product to come out in networking in the last decade, and I think it is the foundation that will carry us forward for another decade.

The first products in our Ethernet fabric offering are the VDX, a top of rack 10-gig high performance switch that costs 30% less than any product on the market, and the VCS, a virtual cluster switch.

The VDX has ports on-demand so you only buy what you need, and then with a software upgrade you can turn it into a VCS that can run the full Ethernet fabric. And if you want to go further, there is a software license to make it fully FCoE capable, fully converged. 

So basically it's a cafeteria, buy-what-you-need plan. You can start with a high performance top of rack switch knowing that, later if you want, you can turn it into an Ethernet fabric. 

How does an Ethernet fabric differ from a traditional data center network?

Traditionally you put in multiple layers of networking and then to get resiliency between equipment in each layer you paired switches and overlaid complex protocols like spanning tree. So the result was lots of layers, lots of redundant switches and probably 40% of the links being inactive because they're in standby mode from a spanning tree perspective.  

What an Ethernet fabric does is eliminate the redundant switches at each layer and the requirement for spanning tree by running switches in an active/active mode that uses equal cost, multi-pathing to enable you to run whatever topology you want. So you can develop very flat, fast network topologies with multiple links between switches that are all active all the time. 

How does this compare to what Cisco, Juniper and, say, HP are pushing?

We're focused on a couple things; unmatched simplicity is first and foremost. There aren't enough qualified people anymore to manage all the complexity. When everyone started cutting costs in the last couple of years, the first thing they cut was people. So the guy that was working eight hours a day is now working 14 hours a day and he's no longer doing one job he has seven. He's not doing them all great and he just can't put anything more on his plate.

Everybody has a list of a hundred projects and the bandwidth for one. So we're trying to deliver unmatched simplicity in the management of the network, the support of the network. Products join the fabric on an automatic basis, they self heal, they actually can be removed from a fabric without operator intervention.

The other thing we're focused on is investment protection. We're not saying take out your existing network, we're talking about making it better. You don't have to rip and replace your current Cisco environment or any of your legacy environment, we just make it better and then we can make it simple and you can grow from there.

So those are the two big differences; if you can make it simple and also protect your investment, it takes cost out.

Now there's one other thing no one wants to talk about. You have data growing at 100% a year, you've got network traffic growing 100% a year, but I have yet to meet a customer that says "My budget's growing 100% percent a year." Good news is budgets are growing, but it's all single digits.

So we have this incredible chasm between budget growth and network traffic and data growth. How do you close it? You can't just say, "Oh, I need four more routers and 16 more switches." What you have to do is provide a new architecture. We have a tremendously different position allowing you to architect for this growth while achieving massive simplification. 

This fabric concept sounds perfect for cloud environments, so let's turn to that topic. How real is cloud computing for the customers you talk to?

When we talk to customers, we say start by virtualizing your own data center, building your own private cloud. Then for the second piece there needs to be some extension technology developed that can logically move the server and the storage together into a public cloud, and you want to be able to move that back and forth at will.

You're going to have to develop this around cloud IDs and so forth and those are the standards we're working on today. We think it's probably 18 months away before you have real hybrid clouds.

But what customers ask is, "If I spend millions of dollars putting this network in, is it going to be able to participate in this hybrid cloud environment?" And we can say with confidence the answer is yes because we've designed this first instantiation of our products with that in mind.

Does the adoption of virtualization and the emergence of cloud represent a new opportunity for companies like Brocade to make inroads against Cisco?

Our research has shown that at any one point in time two-thirds of enterprises are going through some type of consolidation, some type of acquisition or disposition of an asset. That means they are going to have data center issues to consider. Are they going to expand a data center? Are they going to shut one down? Are they going to build a new one? Are they going to consolidate multiple data centers into one? So I look at it as having two-thirds of the market available to me at any point in time that's going to go through some type of transition.

And while the Ciscos of the world don't have to innovate -- can just continue to add functionality and features on their current architecture -- we're willing and capable of dramatically changing the architecture of the data center for the benefit of customers. 

We're basing a lot of the company's future on this Ethernet fabric. When 40% of the links in a data center aren't even turned on but are requirements, being able to eliminate that is a tremendous improvement in simplicity. We can no longer just compete on features, nor can anybody else. But this is an inflection point to put in a new architecture. 

Cisco is obviously trying to shift the playing field with its new Unified Computing System. They have innovated on that side. Are there some things they can do with these computing assets you can't do? 

There is a subset of customers that love a tightly integrated vertical stack. I mean that's what it really is when you take the virtualization layer, the server layer and the storage and you package it all up.

But while the majority of customers like the idea, they don't like the idea of locking in to one supplier. And so we said, "Good idea". However, when you can take an industry standard server and storage, and you can take our BDX Platform with VCS software and create your own. So I can mix and match IBM, HP and Dell servers. I can use EMC, IBM, HDS, Dell, Net App, iSCSI, anything I want on the storage end, as long as you use our VCS architecture you can bolt and create your own.

It sounds like a little more work, except we're going to work with partners and integrators to put together these compute blocks and you'll see some announcements coming out about that.

The issue is we'll be able to build this and be able to deliver it at a fraction of the cost.

How so? Where do the savings come from?

Savings will come from the fact that you'll have choice in servers, choice in storage and from the fact that our interconnect architecture, the VDX product, by definition is 30% less than anything else in the marketplace. 

And then you take a look at the fact that the fully integrated testing of the performance of the product is going to be a fraction of the cost. 

With Cisco going that route and the computing guys aligning with different network players – HP buying 3Com and IBM buying Blade Networks, for example – do you get boxed out to some degree?

You could view it like that or view it as the greatest opportunity on the planet. The fact that I'm one of the few independent guys right now that can sit between all of them and make it work is the greatest opportunity we've ever had. Everybody's creating their own little ecosystems and I can connect everybody and probably do it better than anybody else.

So you're Switzerland?

(Laughs.) I'm better than Switzerland. I'm more like the U.N. But I'll be straight with you. If we didn't have this new Ethernet fabric my hands would sweat. But now that we have a brand new category of product that has positioned us well for the next decade, my hands aren't sweating. 

In fact, when we announce the product we'll have 300 customers testing it already. And when we announce, Q1's supply will be pretty well consumed. 

And that's just out of the gate before we're marketing it. So I have tremendous confidence because it fits so nicely with what customers are trying to do.

Switching gears a bit, where does the convergence of storage and data networking stand in customer's minds? Is that a big priority today?

It's a great checkmark that nobody buys. We've had a converged product for 20 months and it doesn't pay the light bill. It sounds great until you get the storage guy and the network guy in the same room. But we've enabled the technology and allow you to proceed at the pace you want. 

Our new VDX product runs Fibre Channel, 10-gig E, iSCSI, runs NAS so it is a Swiss army knife in terms of protocols you can connect. We've talked to storage guys who want to use it for storage, and networking guys that want it for networking. But when it comes to convergence, I bet less than 5% are actually talking about doing it.

Yet, we've had the broadest FCoE portfolio of anybody in the industry, including Cisco. We've had an FCOE blade for our Fibre Channel directors for almost a year. Cisco still doesn't have one. 

As you look at the competitive landscape, is there anything missing in your portfolio, anything you need from a product or technology perspective to really complete this?

I could always put more arrows in the quiver, I mean anybody could. You always look across the fence and say, "Wow I'd like to have that, I'd like to have." But where we've staked the company's future and where we're really, really good is in very complex data centers. That's a $30 billion market and when you're the size we are that looks attractive. 

So with our new Ethernet fabric, and the brand new high-end chassis we haven't talked about yet -- the fastest high-end networking chassis on the planet that begins shipping next year -- I think we have enough on our plate at this point in time. So no, I don't really need any other piece parts that I can't get through partners.

We've talked about how technology shifts might give you and others a better opportunity to compete against Cisco, but do you see any particular Cisco vulnerabilities?

In the past when Cisco has fallen behind they would go out and acquire a company to catch up. It's going to be incredibly difficult to go out and buy an Ethernet fabric company. In fact, I would say they're going to have to develop this themselves. They do have the piece parts, they have the parts in terms of their product sets, but you don't buy merchant silicon for this one. 

They have to spin an ASIC, develop their own product. They've announced a product called Fabric Path, but they have been a little light on the details. And there aren't competitors in this space; it's the two of us in this whole Ethernet fabric category. So it's going to be a big challenge.

What about a Juniper's Stratus efforts?

Whenever that product actually comes to market. But I think ultimately there will be three companies that provide Ethernet fabrics. It's going to be us first, and if I was a betting man Cisco would be next because they have the Fibre Channel assets to build a fabric, and Juniper will be a distant third. 

You know it took us three years to develop and we had the assets. So it's not something that's simple to get into. But it could be a three-legged stool I don't mind sharing somewhat.

So we're walking around with more swagger, a lot more confidence. Feedback from customers is tremendous. I'd say watch this space.  

Copyright © 2010 IDG Communications, Inc.

The 10 most powerful companies in enterprise networking 2022