Brocade HyperEdge and “Effortless Network”

Brocade moves into the Campus

This week, Brocade has announced “Effortless Network” strategy1 which is their first solid move into the campus networking and replacing the tired FastIron products with the Brocade ICX64x0 switches.

Brocade ICX switches

The first part of the plan is to release the ICX 6430 and 6450 Ethernet switches2 into the entry level space. As you might expect these are Layer 2 Switching with 24 & 48 ports of 10/100/1000Mbps and variations that include 1GbE/10GbE, Power over Ethernet. The price/performance comes form the Fulcrum merchant silicon inside. The ICX products are intended to offer high density edge networking for campus and office networks so large volumes and low pricing is the order of the day.

They also support physical stacking using SFP-based Twinax cabling which neatly allows for vertical stacking in a single rack, or horizontal stacking across the ‘top of racks’ (unlike Cisco’s serial-based stacking in C37xx series). ( See this article on why horizontal stacking is a good idea)

Magic Sauces

Brocade is entering a crowded market niche and one that is being aggressively targeted by HP Networking. Therefore some magic sauce is needed to differentiate this product. Brocade has two sauces on offer via their HyperEdge™ technology as part of their “Effortless Network” strategy.

The “Effortless Networks” feature set that delivers a virtual stacking capability via “Hyperedge” technology. Instead of building physical stacks of switches using proprietary cables, Brocade have created a software features ability that allow switches that are not physically stacked to be logically stacked. In campus networks, it’s problematic to configure hundreds of individual devices, or find a MAC address somewhere in the network. Consider creating a “stack master” from your ICX 6610 in Building H that can control the configuration of all ICX 64xx switches in the building and reduce your operational costs. Hyperedge technology keeps the CLI but reduces the number of physical devices to manage.

The second magic sauce is “Hyperedge” and some explaining is needed. A high end premium L3 Switch supports advanced features such as MPLS or VRF-Lite but the chipset that performs control plane costs more to build and develop. Thus, the first packet in an IP Flow is forward to the control plane for inspection and a decision is made against the configuration on the handling of the flow. The flow entry is then loaded to the forwarding plane for all future packets in the flow inside the switch.

Brocade’s Hyperedge works by using the L3 Switch as the control plane, and the low cost ICX 6400 switches as the forwarding plane for these functions. Thus a traffic flow is inspected by the L3 engine on the premium switch (such as the ICX 6600), a flow entry to loaded to forwarding table in the for all future packets. In effect, you have upgraded all of your L2 money price but needing some careful design. That’s a good deal.

Hyperedge allows for mixed mode stacking of different models of switches for this function. The only fly in this jar of honey is that it won’t be available until 1H2013 and that’s a long way off.

Comment: This is not an OpenFlow play. It’s simply loading the Forwarding Table of the edge switches with control plane data from the a premium class switch somewhere else in the network. Sure, this sounds like OpenFlow, but it’s more correct to say that Flow Management as a network concept has many ways to impact the future of networking.

That’s going to require some careful design of course – but consider that TRILL, MLAG or any other technology still has all sorts of tricksy stuff too……..

Market Moving

Since Brocade has acquired Foundry for it’s Ethernet capabilities, many have pondered when they would extend into the campus. The low end Ethernet portfolio acquired from Foundry was shabby and had a poor reputation overall. This product release shows a long term committment to the Campus Market, and that Brocade is making a serious play.

The us of merchant silicon from Fulcrum Marvel in these units also shows that low pricing and software innovation in networking is certainly possible. This is a development worth watching over the next year as the Magic Sauce becomes real and proven. In the mean time, those ICX 6430 switches start at just USD$1250 a piece. Now that’s good news for a big campus, high end features in a low end box.

Passive/Aggressive Note to Brocade Marketing

There is a time and place for using PDF files but it’s ALWAYS time for web pages that have text on them.

Using PDF files for every single white paper is hugely annoying. I’ll download it if I want to. Further it suggests

  1. that you are anally retentive about your presentation and you want to pretend it’s a printed brochure. Which is quaint but dumb.
  2. that you are control freaks
  3. You’ve forgotten that real people have to read this stuff.
  4. You must hate those people because you make it hard to read your content.

PDF files are optional downloads for people like me. For everyone else, they just don’t bother. If you want people to read your white papers, put them in TEXT FORMAT.

Update 20120310

Not Fulcrum but Marvel merchant silicon. And no plans for MPLS in Hyperedge.

  • John Noh

    Greg, thank you for your coverage of our Effortless Network announcement. I’ll make sure I will get your helpful comments about the collateral to the appropriate Brocade employees.

    John Noh
    Corporate Communications

  • Michael Schipp

    Hi Greg,
        Just a few points, on the layer 3 for the 6400’s, the 6430 is layer 2 only, 6450 ships both layer 2 and base layer 3 (static routes) and is license upgradable to Perm L3 (RIP and OSPF).

    Also of note for the ICX 6430, these switch is fan-less.

    Stacking on the ICX6610 is via 4 x 40G, 2 x 10G for 6450 and ? x 1G on 6430.

    Hyperedge will also support the FCX as part of the mixed stacking as well.


  • Dmitri Kalintsev

    > Thus a traffic flow is inspected by the L3 engine on the premium switch
    (such as the ICX 6600), a flow entry to loaded to forwarding table in
    the for all future packets.

    This sounds very much like MLS that Cisco 5000-series used to have. And it also sounds very DoS-able (intentionally or not).