About the VDX switches
Deni Connor of SSG-NOW had a chance last week to catch up with Dave Stevens, CTO for Brocade on the company’s recently announced VDX switch and VDX futures at the Brocade 2010 conference.
Brocade CEO says Ethernet fabrics will undermine Cisco’s dominance
While we were in Frankfurt, Germany for the global launch of the VDX switch, we shot some video that can be viewed here. In the video, Stevens, explains the new VDX switches and how they work. Following is the rest of the interview.
Next year you have additional switches coming out.
The switches we announced today are top-of-the-rack switches. We are going to do 1GbE VDX switches and larger chassis-based switches next year. We are also going to introduce a core switch that will be used in the data center to aggregate the top-of-the-rack switches at the edge.
Will these switches at some time replace the DCX switches?
No. The DCX switch has been optimized for core Fibre Channel networks. Next year we will introduce the 16Gb Fibre Channel blades for that. The MLX is optimized for high-speed routing environments. The DCX is optimized for virtual machines.
Let’s talk about your converged network adapters.
Our CNAs, which are built on ASIC technology, can run 16 different devices concurrently. There is an important standard coming out the PCI SIG called Single Route I/O Virtualization. The way SR-IOV works is you provide replications of sets of registers on the physical adapter such that any operating system capable of scanning the bus can see any adapters physically configured. So, we’ll configure up to 16 physical adapters on those CNAs. You can configure each of those adapters independently. Those adapters are presented to the operating system, you’d see what look like 16 adapters in a machine.
Let’s talk a little about the channel extensions promised with the VDX switches.
One of the challenges of moving virtual machines between data centers is that people love to do these marketing exercises of let me show you how easy it is to move this VM from data center to data center over distance. The reality is that virtual machines are almost always attached to storage, so if you move the virtual machine without moving the terabyte of storage it is attached to, it doesn’t work anymore. If you really want to get to the point of transparently moving an application, meaning its VM and storage to a remote location, you have to be able to handle the transfer of both of those to the remote site, not just the VM traffic. Today we extend data centers on the IP side with the Virtual Private LAN Service (VPLS); on the Fibre Channel side, we use Fibre Channel over IP to extend the storage network. We are going to take those two technologies and combine them into a single unified platform at the edge of the enterprise. On the WAN side, we’ll offer compression and encryption so the transfer is very efficient; and on the local side, we’ll let users pipe storage-oriented traffic and local traffic into it. We’ll also pick up native Fibre Channel traffic.
More after Thanksgiving.
