The SAN switch maker says it is about to ship what it says is a new category of backbone product that will support multiple technologies – including what Brocade calls Data Center Ethernet, which is a coming storage-friendly version of Ethernet.
The promise of the new backbone devices came as part of a wider pitch Brocade is making for its entire portfolio, under the umbrella marketing label – or architectural blueprint – of Data Center Fabric or DCF. The backbone switches – dubbed DCX devices by Brocade – are the only part of DCF that is new to the company, or that it has not talked about before.
Because Brocade is not giving any details about the DCX devices, it is not clear why they are being called backbone devices or how they will be different to existing SAN directors, other than in size and support for Ethernet.
In an email to Computer Business Review, Brocade described the DCX devices as providing intelligent, high-bandwidth, multi-protocol connectivity and services…throughout enterprise data center fabrics, with plug-in services and intelligence.
Asked how that makes the devices different to existing directors, Brocade said that the devices will sit on the backbone of a data center network and provide switching, routing, fabric and application services to a variety of networks including the server network and the SAN.
The lack of Ethernet support stops Brocade’s current directors from providing what the company also described as the combination of storage networking and server-to-server clustering on a single, converged data center infrastructure.
What Brocade calls Data Center Ethernet is the coming lossless or flow-controlled version of Ethernet that will be needed to support FCoE, which is a protocol that will allow Fibre Channel to be carried over Ethernet connections.All the major SAN suppliers are promising to support FCoE, including Cisco Systems and Brocade itself, and FCoE-freindly hardware is slated to ship in the second half 2008, which means that it will probably ship in 2009.
Brocade like others has argued that FCoE will encourage FC sales, because customers will be able to use FCoE to extend their FC networks out over cheaper Ethernet connections, and that will drive more traffic into the SAN core.
The other Ethernet-borne challenger to FC is iSCSI. iSCSI’s challenge has always been undermined by the fact that is carried on standard Ethernet, which can drop data packets and so may never be suitable for high-end data center storage networks. But FCoE will run on a coming lossless version of Etherne and as a result will not suffer that weakness and will be suitable for high-end applications.
Brocade’s CTO Dan Crain admitted that while FCoE will initially be used as a means of extending FC SANs out over Ethernet, there is no reason why customers could not use it entirely without any FC switches or directors. Since customers want to converge their networks, that means that Ethernet-borne FCoE could entirely replace FC.
But Crain pointed out that the switches supporting lossless or data center Ethernet will not necessarily be cheap – even though the presence of Ethernet in their name might make people think otherwise. We expect that FCOE will cost more than people think – unless some people artificially discount it, Crain said.
Ethernet’s low cost is the result of its ubiquity and huge volumes, but those qualities were not achieved overnight. 10Gbit Ethernet is still very expensive, and it took seven years before 1Gbit Ethernet became free on the motherboard, Crain said.
FCOE standards are still being worked and standards-based products are still at least a year away. The initial value proposition of Data Center Ethernet and FCOE is to aggregate multiple different types of traffic out of the server, so FCoE will start as server connection — and only for those servers where aggregating storage, HPC, and TCP/IP traffic over a common link makes economic sense, he said.
ESG analyst Bob Laliberte described FCoE as clearly one of the fastest-adopted protocols in the IT industry. But he added: What have you ever seen happen quickly in the data center? Large banks don’t just rip out 3,000 ports to replace them with the new thing.
Fibre Channel is not dying. Brocade is positioning itself and its customers for a gradual evolution, not a violent revolution, Laliberte said.
Brocade is just trying to highlight the fact that there’s a multi-protocol world coming. Right now every switch can run FICON at the same time as it run Fibre Channel, but because of data center culture, nobody ever does that. People are already planning what to buy in two years’ time, so Brocade needs to get the multi-protocol message out now.
Our View
The storage industry’s decision to create FCoE was both a death-warrant for FC, and a very smart move that recognized the almost inevitable and actually prolonged the life of Fibre Channel.
If FCoE is used on Ethernet extensions to existing SANs, it will drive more traffic into those core networks, and that will drive demand for larger switches (adding to a pressure for more FC ports that Brocade says is already coming from another phenomenon – server virtualization.)
And even if – in several years time at the very earliest – FCoE makes Fibre Channel switches and directors obsolete, Fibre Channel will live on in FCoE adaptors and in the FCoE protocol itself.
Brocade is making a similarly smart and pragmatic move. It is getting in as quickly as a multi-protocol player, so that it has as much time as possible to cement itself into a new role in the data center, before FC finally fades away.