Brocade VCS Fabric

Just prior to Networking Field Day, the merry band of geeks sat down with Chip Copper, Brocade’s Solutioneer (a job title almost as good as Packet Herder) to discuss the intricate details of VCS Fabric. The videos are well worth watching – the technical details are interesting, but above all, Chip is a fantastic storyteller.

The event was split in four parts:

What is an Ethernet fabric? – a great explanation of Brocade’s view of what a Data Center fabric should be, made even better by Chip’s campfire stories.

Converged Networking and Storage – I skipped this one; there is only so much FCoE I can take and we had a half-day FCoE discussion with Cisco’s Omar Sultan and J Metz a day before.

Multi-Path vs Multi-Chassis – nice discussion of layer-2 multipathing, Brocade’s inter-switch links and load balancing, and the impacts of non-symmetrical network connectivity on traffic flow.

Hard Cores and Soft Edges – given the realities of simplistic hypervisor soft switches, Brocade’s Automatic Migration of Port Profiles (AMPP) gives the networking engineers an interesting toolbox that could make their life easier. Chip did a fantastic job explaining how AMPP works.

More information

If you need to know more about data centers or network virtualization:

And don’t forget: you get access to all these webinars (and numerous others) if you buy the yearly subscription.


  1. Very nice presentations and discussions (3rd video), your first question with regarding to how to distribute traffic over fatter pipes, I think VCS control plane only has knowledge of individual links, i.e. there is no LAG/EthernetChannel pe se inside the fabric, so traffic will be hashed to all available links regardless whether the link is to the same next hop switch or different nexthop switch. Regarding your second question about vPC/vLAG toward split fabric, I don't think fabric itself has mechanism to decide which fabric should take over when this happens, if other side vPC/vLAG device can not detect the fabric failure (most likely they cannot), then there will be a broken L2.
    1. I did not quite get what Chip said that when 2 10G links are in the same ASIC port group, then VCS can essentially make it a 20G bandwidth link and do per-packet load balancing, in this case how VCS control plane is going to treat this 20G link for SPF calculation and how to ECMP traffic over this 20G link?
    2. VCS treats a per-packet trunk as a single physical link. Think about it as a L1 LAG. Any protocol running on top of that will treat it as a single link.
  2. Actually, that's DR. Chip Copper to be politically correct :)

    I just wrote up a long drawn out explanation of the questions above, but decided it was way too much info.

    Let's just say, ECMP takes care of getting from switch to switch and Brocade's fantastic hashing algorithm takes care of the load balancing.

    This technology works. I have this implemented in large production data centers and is supporting tens of thousands of VMs, physical servers, and entire tier 1 applications.

    The only thing I wish Brocade would do is document the increased performance just about all customers have seen using VDX switches in the SAME TOPOLOGY as their older 10g ethernet gear.

    VDX has increased performance 3x-4x on applications and NAS clusters just based on the losslessness and load balancing inherent to the system in the SAME EXACT TOPOLOGY as the older gear with no secret sauce at all!!!! These have been documented by my customers, and it's my customers that are evangelizing this technology for me!

    Next 2 years should be GREAT with VDX!!!
Add comment