Skip to main content

This is not my beautiful network: The false dawn of God-SDN?

For many months now Cisco has been preaching the Gospel about its general disdain for software-defined networking in advance of its Insieme network fabric spin-in announcement that went out on 6 November. Its position is totally understandable, as the primary conceptual difference between SDN and Cisco's newly birthed 'application-centric infrastructure' is that the former does not actually require any Cisco products to function.

But to be fair to Cisco, nobody's perfect (even Beethoven screwed up on even numbers). And the SDN Clown Car has been its own worst enemy, becoming so overcrowded and noisy that the term 'SDN' itself now confuses even the smartest person in most rooms.

So Cisco's knocks on SDN have had myriad easy targets. I also believe that "it's the apps, stupid!" How ironic is it that a charter member of the Open Networking Foundation with commercially deployed, fully integrated OpenFlow-based mixed Optical and Ethernet data centre solutions feels compelled to distance itself from such an iconic term?

Where Cisco seems to go off the rails can, in part, be ascribed to long-term cognitive dissonance.

Starting at the macro level, Cisco calls its ACI solution a "revolutionary network architecture." Show of hands the last time anyone can remember an entrenched incumbent leading a successful revolution against itself? I don't know about you, but the image that springs to my mind is of someone putting a bottle of paracetemol in the toe of a sock and trying to beat himself to death with it.

What Cisco misses in all of this is that this new focus on apps isn't so much about the creation of an 'application-centric infrastructure', but is all about restoring the lost arts of meaningful business innovation and competitive differentiation in its customer base — arts that have been totally crushed by the technological homogenisation of entire industries by Cisco's massive success in the marketplace.

As technically impressive a piece of work as the new Nexus 9000 and ACI appear to be, the real problem is not rooted in speeds and feeds. The 6 November announcement trumpets 60Tbps of switching capacity, 576 40G ports and support for 64,000 tenants - impressive numbers. But the uncomfortable reality is that customers today all have the same templated set of hyper-expensive networking capabilities that will become even more hyper-expensive with the rollout of ACI and the 9000.

But wasn't the Nexus 9000 touted as lowering TCO in the announcement? Absolutely, but one always needs to read the fine print. Cisco says the Nexus 9000 and ACI do not obsolete older Nexus switches and fabrics, such as FabricPath, because "they serve a variety of needs in Cisco's $15B installed base." In plain English: "Not backward compatible."

Do I detect the smell of expensive forklift upgrades in the air?

The way forward is not via a HDN – hardware-defined networking - box-centric proprietary application-centric infrastructure, but through an open, standards-based infrastructure that, quite literally, gets traditional network products – with their glacial R&D cycles – out of the way of applications completely. A great proof point to this is where Pacnet is using the VellOS software and generic white box switches to offer elegant and innovative NaaS services with tiered SLAs across all of Asia-Pacific. No big boxes required.

Already, throughout the relatively brief history of networking technology there have been countless unsuccessful attempts to build a so-called networking 'god-box'. Clearly there's no God software-defined networking controller either, as the problem set is massively larger. There was and is no box – or one SDN 'application architecture' - to rule them all.

Jeff Paine is the vice president of marketing at Vello Systems.