Cisco’s latest “Fabric”: Really?

[#1]

I know it’s been a while since I last had a chance to share my thoughts on this site, but I couldn’t resist this one.  Cisco recently announced its “End-to-End High-Performance Trading Fabric.”  A couple of things about this announcement struck me as particularly interesting. 

First, it doesn’t appear that there is any real news here other than the fact they released performance figures for their Nexus 3068 switch.  It is so unlike Cisco to publicize its performance data; low latency of its switches is not one of their strengths. Cisco has not published performance numbers on the Nexus 5548 and for a good reason. The 5548’s performance on packets sizes of 512B and less is egregious.

Second, and perhaps most amusing, is Cisco’s latest use of the term “fabric.”  Cisco states “the high-performance trading fabric is based on Cisco's holistic data center fabric approach that includes the recently announced Cisco Nexus 3064 and Nexus 5500.”  I am at a loss to understand how connecting 3064s and 5500s constitutes a fabric.  Last summer, Cisco said its strategy for a data center fabric was based on FabricPath.  But neither the Nexus 3064 nor the 5500 work with FabricPath.  Earlier this year Cisco said that its strategy for a data center fabric was multi-hop FCoE.  But the Nexus 3064 does not work in a multi-hop FCoE environment.

This leads me to conclude that either Cisco is confused about the concept of a data center fabric, or they are deliberately trying to muddle the definition of a fabric.

At Juniper Networks, we have a very clear understanding of what constitutes an ideal data center fabric, leveraging the architecture and behavior of the fabric found inside a switch or router and applying it to a data center interconnect.  A data center fabric should provide a flat (single Ethernet look up), any-to-any data interconnect (data plane) while being managed as a single logical switch. This concept forms the architectural underpinnings of our QFabric™ architecture [see “Defining Characteristics of QFabric”] and is the reason why QFabric provides such extraordinary performance (under 5us to traverse the fabric), inherent simplicity and efficiency, all at a lower cost of ownership.


[#2]

Andy, have to agree with you on this one whole-heartedly.  Cisco is doing this for a combination of reasons:

1) There is definitely confusion about what they should be building next, and no clear leadership and direction coming from the committee that is running it.

2) After a consistent string of scathing losses in the financial services market from institutional shops, to hedge funs, quant and HFT, to exchanges they are struggling to find their own relevancy - and thus turn to merchant silicon weeks after John Chambers says 'it has no value and anyone an sell it" and also use a silicon architecture that is already being obsoleted by their forthcoming Nexus 3064PQ-10, and then being obsoleted again in Q1 with the PHY-less version.  This rapid churning of hardware is indicative of a rushed product line and knee-jerk reaction to competitors.  Lastly, the silicon in question is not even the lowest latency silicon available for a market that views low-latency as the top purchasing criterion.

3) Cisco has not managed to integrate their vertical solutions focused people into anything other than marketing. So what we see is people with the title 'solutions engineer' sit down like a kid with a Lego set and try to build the best plane/car/boat they can with Lego bricks but find they are missing a few critical pieces that would make the plane/car/boat just perfect and impress all the other kids and parents.  Until someone makes some structural changes, the best we can hope for is this marketing blather, no real products focused on the financial services market or fabric architectures as described above are likely to appear.

dg