Brocade has been building network fabrics for a long time. The company is currently on its sixth generation of Fibre Channel fabric, and since launching its Ethernet fabric two years ago the company claims to have 1,200 customers with over 200 devices deployed.

But the shift in focus to fabrics actually undermines Brocade's core business model. The company has traditionally made its money selling data centre switches, but fabrics provide more capability at a lower cost.

This may seem like an odd strategy, but Brocade's new CEO Lloyd Carney said that the company is replacing its own product with a more cost-effective solution because it knows that fabrics represent the next wave in enterprise networking.

He said that if a customer was standing up a 200-port environment, they could spend $480,000 on Brocade's standard switches, compared to $800,000 from its main competitor, or it could spend as little as $250,000 on a fabric that provides more capability, more flexibility, and is easier to manage.

“If you want to add another switch into this infrastructure, you don't have to go and upgrade spanning tree tables, you don't have to worry about routes, you just plug it in,” he said in a keynote speech at the Brocade Vision executive forum last week.

“We self-learn, we load-balance across the ports, we choose automatically the best path to run the traffic and keep track of the traffic. This is a best-of-breed solution for cloud environments, for large data centre environments, and that's why we are eating our own young.”

Market challenges

Carney, who has previously held a number of high-profile networking positions, including chief executive of network management software firm Micromuse and chief operating officer of Juniper Networks, was appointed as CEO of Brocade back in January 2013, replacing Mike Klayko.

Carney told Techworld that, coming into the company, he knew that Brocade had great technology, both on the traditional networking side and the storage side. However, the best surprise he has had since joining is discovering how thorough the customer base is.

“Every time I ask about a customer and say we should be in this customer, the answer is we're already there,” he said.

“If we didn't get one more new customer – all we did was satisfy the customers that we have, make sure that they're delighted with us, make sure that we provide them with stellar support and service, and sell them the other products in our portfolio – we would be wildly successful.”

The company's list of customer references is indeed impressive, ranging from web giants like AOL to universities, banks, government departments and research facilities such as CERN. Carney described it as “a who's who of the IT world”.

However, he admitted that Brocade has traditionally not been very good at marketing its products – particularly its Ethernet fabric. He said the big challenge will be to show customers how efficient an Ethernet fabric is to operate, and how much money they could save by deploying one.

“The push for the next two years is letting people realise that we're not just Fibre Channel experts, we're storage connectivity experts, and we're really data centre connectivity experts. You want to connect your server to storage in a data centre, we are the guys you should talk to,” said Carney

The UK is in fact the country in which Brocade's Ethernet fabric has the highest penetration of anywhere in the world. Carney said that the UK has been an important proofing ground for the technology, and could lead the way for greater adoption in Europe, the US and Asia.

Network convergence

One of the topics that Carney is asked about most frequently is network convergence, (providing fibre channel and Ethernet over a single cable). While this makes sense in theory, allowing customers to pay for just one cable rather than two, in practice it ends up being more expensive, according to Carney.

This is because the standard today is 10G Ethernet, and over that 10G Ethernet a customer is lucky to get 4-6G of Fibre Channel capability. At the bottom of the networking stack, however, there is 16G going into the core device, and Carney argues that it makes no sense to try and drive a 16G infrastructure with 4G of fibre channel.

“Where converged makes sense is if, at the server level, some of these servers don't need a full 10G pipe, so you can have 4G of fibre channel and 6G of Ethernet. But once you get to the bottom, you need real fibre channel, because it's just too much traffic,” he said.

“If you do it end-to-end it costs you more money than using a traditional infrastructure, and you have less performance, because the best you have at any one of these cables end-to-end is 4G to 6G of capability, and you've got a 16G interface at the bottom.

“It's like you've got a swimming pool, and you can either use a hose from the garden or a hose from a fire truck to fill it. What are you going to do?”

Carney said that server virtualisation is making network convergence even more impractical, because multiple 10G links are needed just to feed the Ethernet coming out of servers. By the time 100G is cheap enough to become the standard, these servers are likely to be running 200 virtual machines, and they will chew right through 100G.

“The further you get from the servers, the more convergence makes no sense. The only possible place it makes sense is at that low-power under-performing server. If you have high-power servers with a bunch of VMs on it, it doesn't even make sense there. And that's what customers are finding out,” he said.

“Try it, stand it up, have your guys test it and see what happens. Invariably they come back and say the performance really stinks, I get bottlenecks all over the place.

“You don't just plug it in and all of a sudden it knows to do a certain amount of bandwidth for fibre channel, a certain amount of bandwidth for Ethernet. You have to program it at every step of the way to do that, and the guys are like, forget it.”

Software-defined networking

Software-defined networking (SDN), on the other hand, is a big area of focus for Brocade. Last month the company added hardware-based support for the OpenFlow SDN API and protocol to its NetIron and MLX series 100Gbps routers.

Brocade's SDN strategy also includes the company's existing VCS data centre Ethernet fabric technology, which Brocade says can facilitate SDNs through VM awareness, active-active topology resilience and optimisation for virtualised environments.

Carney said that a lot of the attention around SDN has happened because of VMware's acquisition of Nicira last year. In particular, he said the Open Daylight project – which includes the likes of Brocade, Cisco, Juniper Networks and IBM – is a bid to ensure that VMware does not gain the same advantage in SDN that it has in the hypervisor.

“SDN got pushed into the limelight because VMware spent $1.2 billion to buy 40 engineers working for Nicira. If that acquisition didn't happen, we wouldn't be talking about SDN,” said Carney.

“It would still be happening, because it's been driven by economics and physics, it just makes sense. Intel has been pushing SDN now for maybe seven years – nobody knew about it, no one paid attention to it, it was just happening. It makes sense only because of the scale and capability of the Intel processors, and the dropping memory costs.”

Carney said that adding memory into one of the large switching vendors' boxes costs ten times more than adding memory into an Intel processor, and the performance on an Intel processor is every bit as good – if not better – than the processors used in a lot of stand-alone networking devices.

“For Intel it's just a straight market play. How do I get more of the data centre? So anything in the data centre, if it's software, why doesn't it just run on one of these cores? That's been the driving force all the time. No one ever paid attention until someone popped a billion dollars plus for a little company.”

In spite of this, however, Brocade does have a team focused on building apps around SDN, as well as customers playing with it. Carney said that SDN is an important extension of Brocade's fabric strategy, because it will eventually provide a better solution for its customers.

“Our fabric is SDN-ready. The guys designed it with SDN in mind, and that's why we made the acquisition of Vyatta last year. Vyatta is working on our SDN control plane. They brought to us virtual appliances they brought the virtual firewall, they brought the virtual router with them, and that is absolutely a path we see people going down,” said Carney.

“In the world that we live in, there are no revolutionary changes, because you're running real businesses, you can't afford to do that. There are evolutionary changes, and you'll see the partners will go to SDN, functions will use SDN, and then you'll get bigger organisations, more mission-critical functions using SDN, and then you'll get overall conversion over time.”

Integrated systems

One important shift for Brocade is the way that enterprises are buying data centre equipment. Rather than buying servers, networking equipment, storage and management software separately and then assembling them into custom configurations, they are increasingly buying them as a single package.

IDC predicts that sales of integrated infrastructure and platforms will reach $3 billion this year, and increase by 50 percent each year for the next five years. Meanwhile, sales of traditional data centre hardware – purchased unconfigured and unintegrated – will increase at a rate in the “mid-single digits,” according to the report.

Many major vendors already have products ready for this market. For example, VCE, the company spun out of a consortium of Cisco, EMC and VMware, offers integrated compute/network/storage bundles called Vblocks. IBM, HP, Dell and Oracle also offer their own integrated systems.

While this trend may seem like bad news for pure-play technology vendors, Carney sees it as an opportunity. By working with IT integrators that specialise in providing end-to-end infrastructure solutions, such as Arrow ECS, companies like Brocade can have the best of both worlds.

“We have a stack with Hitachi, we have a stack with Fujitsu, we have a stack with EMC, and we have partners who are building stacks using our product – one of the partners said to me, you guys are like Intel Inside,” said Carney.

He said that Brocade prefers to work with partners that are building solutions, and tailoring those solutions to specific industries, rather than simply flogging equipment.

“The partner who I pay attention to is the partner who comes in and says, I am a specialist with universities or I am a specialist with insurance companies – here are the applications they run, I want to create a stack for their applications and I want you to be a part of that solution.”

Carney claims that putting Brocade's network fabric in a stack makes it easier to deploy and easier to manage. The customer also needs fewer resources, so it's more cost-effective up-front, and they need fewer people to manage it over time, so they save money on an ongoing basis.

He added that big monolithic companies that offer pre-integrated multi-purpose stacks may have difficulty winning customers, because most CIOs prefer to work with partners that have an in-depth knowledge of their industry and who speak their language.

“We think it's a big opportunity for us. Any time there's change, if you're smart about it, there's an opportunity. Change is good. Steady-state causes lock-in,” said Carney.

Building data centres for the cloud

Ultimately, of course, the customer couldn't care less what is in the stack. All they care about is whether it works and how much it costs to run over time, and this is driving the shift towards hosted services and cloud computing, according to Carney.

He admitted that, even though Brocade is one of the more qualified companies in the world to build and run its own data centre, it still uses a cloud service for its CRM, because it's more cost-effective, and it uses a hosting provider for its redundancy.

“That's the bifurcation you're seeing. You're seeing companies that really have to have their own IT infrastructure build it – MI5 is not going to outsource to Amazon – but even the US government is outsourcing some of their non mission-critical things,” said Carney.

“That's a trend that we're embracing, and our focus is on building data centres for those people, making sure they can build the most efficient, most scalable data centres around, so that's the path we're on.”

Find your next job with techworld jobs