Brocade: Expect A Chassis Fabric Switch

Brocade emerged from the storage area networking (SAN) field to become a bigger network player. It is active in the cloud, and is taking on Cisco’s Unified Computing System (UCS), though without moving into the sale of servers. Instead the company, which acquired Foundry Networks for $3 billlion in 2008, is pushing its “fabric switch”, a flat lossless Ethernet for data centres.

Brocade had a Tech Day in London this month – and TechWeekEurope met Marcus Jewell, Brocade’s head of UK and Ireland. He spoke to us about fabric switches, and the competition – along with some remarks about the channel, which are over on our sister publication ChannelBiz.

Stand by for a fabric chassis

Did you really invent the fabric switch?
It’s a bit like cloud. Did we invent the SAN market? That is a true enough. We define a fabric as a flat lossless network, so therefore we believe strongly that we were the inventors of flat lossless Ethernet.

The only way you can measure these things is who has references. We have 560 active users of fabric, bearing in product has only been shipping for twelve months. That’s a pretty good ramp, and we will shortly be expanding that product range. The whole Ethernet fabric product range will move to the next generation in the next six months.

What is coming next?
Old habits in the market die hard, and there is a requirement for chassis switches, alongside the fixed configurations we have at the moment.

Our chassis based fabric solutions will incorporate Layer 3 functionality. People like to see big boxes with lots of blades, redundant modules and redundant power supplies.

There is no point trying to change the world overnight if the client wants that kind of solution. You could argue that you don’t need it. You can fit fixed configuration switches together. But it is important that we meet client requirements. There is no  point making a product no one will buy.

So even if you disagree you would make the product?
Oh, I wouldn’t say that. There are some advantages in the chassis in terms of the amount of management, the quality of the ASICs – the board power is higher – and the density is clearly better, which is important because you use less racking space.

When you connect individual products together, you have to give up some of those ports to do the actual connectivity. With a backplane you don’t lose that, so there is a benefit to chassis. And also you are only doing the industrial design for one box. There is less tin in it and more upgradeability.

You might want to mix and match port types, for instance combining 100G with 10G. You might want to have single-Gig, but hardly anybody does any more in the data centre. You might want Layer 3, you might not.

What about the competition?

Fabric-wise, Cisco and Juniper aren’t yet doing anything that worries you?
Of course we would be foolish to say that. Cisco has a solution called FabricPath, based on their Nexus data centre switches.  We don’t believe this is a pure fabric. It is not what we would call a flat Layer 3 network.

HP haven’t climbed high enough into networks?
No, HP is mostly campus, what we called edge switching in the good old days, and now we call it campus. And the H3C (Huawei 3Com) acquisition includes data centre products, and they have a completely different OS. They are a completely different product set. They are not really integrated, and they don’t have any fabric technology.

They talk about it but they don’t have any. They’ll get there because of the sheer scale of HP.

You don’t feel the need to do closer integration with servers, UCS-style?
No, we focus on networking as our single strategy. The best unified computing block out there now, is clearly Cisco UCS,but it isn’t something we feel will prevail in the market, because we feel closed architectures never do.

Best of breed is the best way forward, but we do provide a lot of close integration with virtual compute blocks – for instance with Fujitsu and NetApp, and also with IBM.

It is hard to integrate into a proprietary based system. I think history has shown that you end up paying 25 percent more over the lifetime of the product. We say, if you want a closed system, why not buy an IBM mainframe?  If you say that to an IT director they say – are you crazy? But buying a closed compute block is the same thing.

There is a big missing part in Cisco’s UCS, and that is storage.

If you are being best of breed, what are you missing out on?
As the networking market changes, there are always things to look at. We have a strong wireless business, if we wanted to extend that.

How much do you know about cloud computing? Find out with our quiz!