◎ Brocade’s Data Centre Ethernet Strategy

I attended the Brocade Analyst and Tech Day last week as a guest of Brocade where I got to learn more about product, technology and strategy. In particular, the event was led by the launch of the Brocade VDX 8770.

TL:DR version: I have a better understanding of Brocade’s market strategy, insight into the technology and my current view is that Brocade has a good product here. The proof will be in the delivery, and whether Brocade can let go of its storage legacy (Fibre Channel) and properly commit to Ethernet. They will need to convince networking professionals that their product managers understand the market and requirements to get them to switch to Brocade. At this time, I think Brocade has a chance of making that happen based on the emphasis of VCS Fabric and Automation.

Where I’m less comfortable is that Brocade will rely on external parties to deliver the software automation – that’s a strategy that has NOT worked in the last 20 years. There is no reason to believe this anything has changed.

Brocade VDX 8770

The primary focus of the event was the launch of the VDX 8770 switch. So lets start with some action photos because Big Ass Switches are always cool. I like software, don’t get me wrong, but I have  special place for big metal boxes.

Click to Embiggen

BIG METAL BOX!! Yay.

BIG FANS for the line cards and fabric card at the top and fan outlets for power supplies below.

The next view has better lighting and you can see the six fabric slots in the vertical center with four payload slots on either side. The dark areas are the air inlets. Under that is the management cards, then power supplies.

Eight power supplies !! (expect you don’t need them all, and may never need them all.

Initial Opinions

My basic opinion is that  the Brocade VDX is impressive on “PowerPoint” with the following specs:

  • Two models – 4 slot / 8 slot chassis for the data centre
  • At launch the line cards available are:
    • 48 x 1GbE
    • 48 x 10GbE
    • 12 x 40 GbE
  • All of these cards are line rate, non blocking.
  • Backplane has theoretical 4 Tbs or more of raw bandwidth per slot and therefore Brocade are targeting 100Gbe density in the future.
  • Chassis is designed to have a long use life of five to ten years (hence 4Tbps plus )
  • Can support 8000 ports in a single fabric up  48 x 100GBE per slot (4Tbps backplane, fabrics & line cards are obviously the future. Headline numbers depends on modules used etc etc, but quick maths on 24 chassis with 4xSpine / 20xLeaf using 8 slot and it’s reasonable.
  • Use multiple chassis to build a single fabric of up to 8000 ports in a single fabric.
  • L2 Multipathing via TRILL and VCS.
  • L1 channel bonding at PHY layer (Brocade trunking)

Key Features

Brocade wants to focus the marketing message on three key value propositions. Expect to hear a lot about these topics.

1. Multi-Gateway L3

  • Every node in the fabric can become a L3 gateway. improves utilisation.
  • Delivering four L3 gateways sharing a single MAC address in the initial version. 4 chassis makes a nice L3 enabled Spine design for early adopters.
  •  improved performance – no L3 bottlenecks , routing at source.
  •  improved availability – redundant routers etc etc
  • east west communication becomes more viable.
  • L3 enable routing on just a single rack. (I don’t remember why I wrote this, maybe it will come to me later)
  • Edge forwarding for L2 and L3 for ingress and egress.
  • “Brocade trunking” will load balance at L1. Evenly balanced compared to LACP. Interesting design options.
  • Complete Multilayer multipathing at L1, L2, L3.

2. Scale

  • Brocade builds their own chips. Optimise the chip performance
  • increase the size of the MAC table – 384 000 MAC addresses
  • TCAM supports 384000 IP routes
  • reduced latency of the port to port communication. 3.5microseconds port to port anywhere in the chassis (much less than other products)

3. Investment Protection

  • recurrent message
  • future silicon will includes the ability to support tunnels.
  • Chassis can build from standalone to fabric to large scale fabric. Combine with VDX6xxx to do top of rack and minor cores as needed.
  • Build from pair of VDX8770 and scale up as you grow.
  • Remember technologies like AMPP (check out part 4 of these videos from TechFieldDay Virtual Symposium.
  • Chassis designed to last through many growth phases.
  • inspecting the tunnel to gain visibility into the VXLAN tunnels in the current silicon. VXLAN in next generation silicon.

Commitment to Overlay Protocols

Brocade has gone public on a commitment to overlay Protocols such as VXLAN in 2013 for the VDX and MLX products. here is the slide making that commitment:

As noted here the Brocade ADX already supports VXLAN termination. More on this later.

Market Strategy

So if Brocade has a good product, then it’s worthwhile considering whether they have a viable software and market strategy. You don’t buy the software as one-off item, you make a commitment to software platforms over a period of years and you got to _want_ that software (something that Foundry never understood). I’ve captured a number of photos from the presentations that I think show the essence of their approach.

Firstly, committing to Fabric and SDN.

This SDN and Fabric commitment was repeated a number of times by the management and then down into the technical meetings I had later. Get the feeling it’s serious.

Quick look at other products in the VDX data centre family. The NOS 3.0 software adds a lot of features to those existing switches too.

I could map an architectural strategy around this – cloud, SDN, orchestration, automation.

Repeating again the commitment to APIs, SDN, abstraction such as VXLAN and so on.

 

This slide a good wrap on Brocades sales focus. Automation at many levels (internal/self and external). Efficiency and technical excellence as differentiation to other vendors — interesting approach in that Juniper is also attempting this. Cisco isn’t striving for technical excellence to date and instead focussing integrating with legacy equipment and maximising value of older equipment.

Oh, in case you missed it. Brocade will thump the drum on having their own silicon at every chance they get. They are well pleased about their silicon design and probably with good reason. The silicon is long way ahead of Cisco and an advance on Juniper QFabric. 

Brocade ADX

I went to the technical area later to learn more about the ADX (not previously known to me). Turns out that the ADX is the Switch Fabric with 3 xFPGA connected to the inputs therefore it’s mostly a software/firmware controlled platform that has a lot of flexibility. The use of FPGAs means that some updates to the GA code libraries adds support for VXLAN in short order.

The current version of the ADX supports VTEP today.

The latest version of the ADX also support contexts for multi-tenant applications but cannot perform VTEP inside of these contexts. I cannot confidently assert that this is on the roadmap as the conversation got a bit fuzzy at this point. The ability to VTEP in a secure manner with separation to meet PCI guidelines is key to deployment.

Software Automation

There was much ado about “automation” and “orchestration” which appears to be terminologically congruent with “SDN” and “OpenFlow”. I can broadly accept all of this being lumped into single “SDN” term. For Brocade, SDN is about supporting OpenFlow and other standards APIs for configuration. And SDN means working with “partners” to deliver a software layer above the hardware via an abstraction layer.

I’m not sure that this works. Focussing on making hardware sounds like a good idea but waiting for someone else to add value to your product using software hasn’t been successful in the last 20 years. Consider SNMP managers such as HP OpenView (comedy laugh), BMC Patrol (stop me now) and the overall lack of capability. While Solarwinds has managed a decent platform, it took 20 years to arrive.

The EtherealMind View

Brocade is off to a good start. The VDX 8770 rounds out the product set for the DC and looks good in PowerPoint. Now, the delivery and whether other people agree is not so clear.

My contact with Brocade field engineers and sales is less than inspiring because they are almost all storage people. Networking people don’t want to talk to storage people – they have different perspectives and views i.e. all wrong. And leveraging off their storage expertise is not necessarily a feature, it’s mostly a negative at my level of the working stack.

I like the VCS approach to fabrics – automated, simple, self configuring and a number of extended features. The hardware products looks good and probably even great over time.

Lets see how the delivery goes. The proof, as they say, is in the pudding and we have pudding, here.

Disclosure

Brocade paid for my travel and accommodation to attend this event. I didn’t receive any other consideration or control over my content. Brocade neither asked or expected me to support this event or to write about their products or company.