Another Missile Fired in the Cisco vs. VMware SDN War

Network Virtualization

Yesterday I was reading about Cisco’s fourth quarter earnings results, as you do when you are bored and waiting for the next episode of EastEnders to start—well, we all have to take a rest from SDN goodness every now and then. Now, this was interesting for two reasons. It was the last quarter under the leadership of big bad John Chambers and the first announced by new head honcho Chuck Robbins (sounds like a cross between a cage fighter and a liberal comedian). Firstly, congratulations are in order on the results—Cisco exceeded analysts’ predictions of $12.6 billion in revenue, with $12.8 billion and a per-share profit of 59 cents, up almost 4% over the previous year, and an overall year-over-year increase of 4%.

However, this is not the crux of what I am going to talk about. As part of the earnings announcement, Mr. Robbins threw an SDN grenade at VMware and its NSX product. Part of his statement resonated with me—soothed that nagging itch on my back that cannot be reached any other way than by rubbing one’s back up and down a tree trunk, grizzly bear style. He stated that Application-Centric Infrastructure (ACI) is a better fit for the enterprise because it can scale beyond and into the rest of the network.

What I think he meant here is that ACI, with its physical presence in the form of the Nexus 9K series of enterprise switches, allows Cisco to easily roll out its SDN overlay to physical devices in the network. You know—bare metal servers, things running physical workloads, HPC environments, non-x86 workloads (mainframes, PowerPC-based AIX machines, legacy SPARC servers, AS/400s, Teradata). Yes, those things do still exist in real-world data centers outside the ivory tower that is Silicon Valley, and I am including FTSE 100 and Fortune 500 corporations in this list. ACI can interact with these devices in a way that NSX never can, with its virtual-centric view of the world. For NSX to interact with physical devices, it must break out of the SDN overlay and return to the physical paradigm via a construct called the VTEP. Now, these can be virtual—i.e., hosted on an ESXi host—or as is becoming more prevalent, physical—i.e., built into a partner’s switch or routing device. Examples include Juniper’s MX Series, Arista’s 7150S, and Brocade’s ADX Series. Both approaches have merit. VMware’s is a true software construct that fully resonates in a software-defined data center. Cisco has taken the much more pragmatic view that legacy will always be there, so let’s be inclusive and include them.

Much as I love the elegance of VMware’s NSX solution, I am becoming more and more aligned with Cisco’s viewpoint on implementation. VMware’s reliance on breaking out of the environment will lead to pockets of SDN goodness surrounded by oceans of physical legacy networks, feeding those environments that cannot, for various reasons, be virtualized.  ACI can natively extend down to the physical device, and as soon as the traffic from that physical device hits the switch, it is smothered in SDN goodness. Cisco can even drop down to the virtual level due to its Nexus 1000V series of switches, thereby providing end-to-end overlay. To me, Cisco’s position is more akin to the ocean than an island, and we have all seen the damage that water can do to land.

"SDN" damage done to the the Norfolk UK coastline by sea erosion.
“SDN” damage done to the Norfolk UK coastline by sea erosion.


Posted in SDDC & Hybrid CloudTagged , , , , , , ,