I had a briefing with Brocade soon after I wrote the article; The Server Vendors v Cisco: Is This A New Technology War?. Brocade was a little aggrieved that I chose to use Cisco’s Nexus 7000 switch to illustrate the idea that the network hardware vendors were now in a stronger strategic position than the traditional server vendors, in respect of corporate computing. With the Nexus 7000, Cisco was providing technology that runs an OS and manages the whole network, virtualizing the connections between servers. It is thus in the ideal position to manage service levels.

In such an environment servers can be viewed as “application execution resources” to be used as necessary – rather like micro-controllers in a PC. This is especially the case when you introduce dynamic virtualization into the environment, where new instances of virtual machines are created “on the fly”. (If you’re not sure what virtualization is click here.) When that happens, dynamic management of the bandwidth between servers is necessary.

In any event, Brocade is quite correct, I could have used Brocade’s DCX Backbone to illustrate the point that I was making. For the record, the spec of the DCX Backbone is impressive. Here’s a partial list of the product dimensions:

• 896 x 8 Gbit/sec Fibre Channel ports (in the 2 chassis system)
• 12 Terabit/sec of system I/O bandwidth (yes that’s 1.5 terabytes per second)
• 99.999% availability
• Energy efficient: One-half Watt per Gbit (10x more efficient than competitive offerings)
• No over-subscription
• Explicit consolidation and virtual machine support.

Brocade is claiming leadership in virtual machine support, core bandwidth and power saving, and the technology is available now. (Cisco wont be delivering the Nexus 7000 for a few months yet).

The early use cases of the DCX are in two areas:

1. Server consolidation/virtualization. It isn’t often mentioned in the various server consolidation studies I come across, but there’s a potential gotcha in server consolidation. You can use VMware to put several virtual machines on the same server, but the network bandwidth might not be adequate for the applications that will now run from the same place. The problem is worse if you’re hoping to deploy virtual machines dynamically. The DCX makes this problem go away.
2. SAN upgrade. With I/O bandwidth at 1.5 terabytes per second, you just know there’s not going to be a problem throwing moving even huge heaps of data around.

The Heart and Lungs of the Network

In the briefing, Brocade preferred not to talk in terms of “now the network really is a computer.” I attribute their reluctance to embrace such a vision to the fact that their primary partners are the server vendors. No point in alienating your friends. Instead Brocade proposed the idea that the DCX could and would become “the heart and lungs of the network”. This is a reasonable metaphor, as far as moving processing and data around the network is concerned – although it is a little data-centric.

Architecturally, we had been confined to building applications to run in specific collections of server resources. With the DCX, the relocation and replication of data is largely unconstrained. In other words, some applications are constrained to some degree. Those with heavy workloads may need to be confined to specifically configured server environments for the sake of load balancing and failover, but most applications will not be constrained in this way. The DCX gives you the flexibility to move just about anything to any server or virtual machine and adjust the network to cope with it.

Aside from its contribution to SAN consolidation, network consolidation and server virtualization, I see the “network in a chassis” which is what the DCX effectively is, as the antidote to problems that have not yet arisen. Two particular issues will soon mature. One is how to support converged communications (VoIP ++) within the corporate network in a way that guarantees the 99.999 service that all communications services mandate. The other is how to cope with passing big blobs of video across the network.

System Management for the Network in a Chassis

Brocade provides four distinct management products for the DCX, all of which relate to SANs. They are:

• SAN Health: Diagnostic software that checks for SAN problems.
• Fabric Manager: A SAN Management tool for Brocade FOS-based SANs.
• EFCM: Which provides a single console for managing multiple SANs of all varieties.
• An SMI-S Agent with SMI-S interfaces for third party OEMs

Brocade provides management tools for the network “data fabric” but intends to leave other aspects of managing this environment to the system management vendors such as CA, BMC, EMC, IBM, HP and Sun. There’s sense in this. System Management is not Brocade’s business – at least not yet.

This is a posting in the Virtualization Focus Series. Click here to see an index of such postings.