VMware’s SDN Strategy is No Threat to Cisco, says Mike Fratto

August 8, 2012 - 3 Comments

For those of you wondering about the impact to Cisco of Software Defined Networking and the combined SDN strategy of VMware and Nicira, I point you to a very rational and well-articulated article by Mike Fratto of Network Computing, that basically says Cisco doesn’t have much to worry about. (Enterprise Strategy Group had already said something similar, by the way).

Specifically, Fratto says:

Mike FrattoThe lack of programmability in existing networking hardware is certainly a problem, but VMware’s acquisition of Nicira does not mean that Cisco and its ilk will be marginalized… It does mean the role and management of the physical network is changing, and I think Cisco is further ahead than most of its competitors in creating a vision for the next phase of networking.

I couldn’t agree more. Since Cisco live! when we announced our Cisco ONE strategy for network programmability as well as the advances in our Nexus 1000V portfolio for virtual network overlays, I have been posting on many of the same points.

My take here was that the VMware-Nicira acquisition did not portend a strategic break with Cisco, and while there are some obvious overlaps in our product lines, there are still a number of areas of collaboration, cooperation and interoperability. The virtual network infrastructure is just one piece of a larger software stack and the differentiation will likely be decided in the orchestration, management and applications built on top of the newly programmable infrastructures sometime down the road. Fratto goes on:

Let’s review what Cisco offers in the data center. Its go-to product line in the data center is the Nexus family of switches, from the core with the Nexus 7000 to fabric extenders that reach to the top of the rack and into blade chassis. The Nexus 7000 and 5000 run FabricPath, a multi-path Ethernet protocol similar to the IETF’s TRILL, which lets network admins create multiple loop-free paths through networks.

Cisco also has the Nexus 1000V, which has replaced WMware’s vSwitch. Cisco has continually added features like vPath to the 1000V, which can be used to add Layer 4-7 services like load balancing and firewalls its virtual switch. Cisco is also supporting VXLAN, which enables overlay networks similar to Nicira’s STT. Finally, the 1000V will be available for Windows Server 2012 and Cisco has demonstrated the Nexus 1000V on OpenStack at Cisco Live. The demo showed the 1000V running as a virtual switch and interoperability between Quantum (OpenStack‘s virtual networking project) and Cisco’s Virtual Supervisor Module.

This is a great summary of what I’ve been pointing out over the last two months, particularly about vPath as a strategic enabler for Cisco, the importance of VXLAN (which we co-designed with VMware and others), and how OpenStack on the Nexus 1000V will open up network programmability for applications on our virtual infrastructure.

Oh, and speaking of SDN and network programmability…

Cisco also announced its SDN strategy at Cisco Live with Open Network Environment, which includes an SDK and API called onePK for routing and switching platforms such as the Integrated Services Router G2, Aggregation Services Router, Cloud Services Router and Nexus data center switches. Cisco says onePK will start shipping by the end of the year.

With Cisco selling physical and virtual networking, integrating the Nexus 1000v with multiple hypervisors and writing a module for Quantum support, does it really look like VMware/Nicira poses much of a threat to Cisco? Not from where I sit.

Well said. All the stuff I’ve been beating around the bush about, Fratto is pulling together very succinctly. Fratto wraps up with how he thinks this product strategy will play out for VMware:

Here’s what’s going to happen: VMware is going to use Nicria’s technology to provide a management framework within vCenter for virtual and physical network equipment and because of VMware’s market presence in the data center, all the networking vendors will work with it. Just like with storage, the VMware management components will be a subset of functions needed to provide networking for VMs, such as defining connections and profiles. More importantly, however, you’ll be able to easily insert within vCenter services like load balancing and firewalling into the connections.

VCenter is going to assume that the network can fulfill the requests made by virtual machines. It will be up to vendors and network engineers to design and maintain robust and reliable networks that meet that assumption.

And that’s just what Cisco as a vendor plans to do. We still believe it’s important to tie into the VMware management infrastructure as well, as we support both vCenter and vCloud Director for multi-tenant data center provisioning. As well as our own Intelligent Automation stack. Again, Fratto has it right.

Meanwhile, we still get questions on what is really important about Cisco’s virtual network overlay infrastructure. To which I respond: Nexus 1000V is a very proven environment with over 6,000 customers as of July 2012. Nexus 1000V provides management and feature consistency with Nexus physical data center switches for a seamless integration of physical and virtual and cloud environments. Cisco also provides much more in terms of virtual network services, and a complete virtual control plane than any of the other competitive alternatives. We also integrate data center network features like LISP and OTV on top of our virtual networks, for greater scale and application mobility, e.g.

Continue to watch this space for more deep dives and perspectives on Cisco ONE and the Nexus 1000V as this evolution plays out.

In an effort to keep conversations fresh, Cisco Blogs closes comments after 60 days. Please visit the Cisco Blogs hub page for the latest content.


  1. It is a very interesting landscape forming in the world of virtual switching/content delivery. This battle still isn’t won by any vendor. If VMware had it right, they wouldn’t have brought in the likes of 1000v a few years back or acquired Nicira. This is a growing space and seems to be a charge for the opportunity by many of the market leaders.

    As stated, the physical network isn’t going any where. The largest shops (Google) have found ways to curb costs by engineering hardware and running it on an in-house stack and openflow. These are exceptions to the norm. It takes a lot of routers and switches to make this model justifiable.

    There are a lot of frienamys coming out of current computing environment and especially in this growing market.

  2. While I respect what Cisco is doing with Cisco ONE and Mike Fratto’s view that Cisco is not threatened by VMware’s acquisition of Nicira, I would beware of any complacency relative to technologies that are aimed at simplifying network deployment, operations, and use; better utilizing network resources (and connected resources); and improve service quality via heightened netwok dynamics. Most of the sure-bet IT advancements (Cloud, SDN, BYOD…) are driven by these same Top 3 benefits – simplification, utilization, and adaptation.

    Assuming that Nicira is going to be relegated to just another management system echoes the same assumption server vendors made when first examining server virtualization technology. There is a big difference between simple management and complete control. Network virtualization and SDN solutions to come will be far removed from VLANs, policy managers, multi-service devices, and virtual switch complexes of today. Here, Cisco needs to remain in complete control of the network – the entire network.

    • You are right Mark. I think the context of Fratto’s piece in particular was all the noise about SDN commoditizing the physical network and being a real threat to Cisco, and he wanted to counter that. Clearly we agree, but nor is it the case that Cisco is ready to get complacent about this. We are keeping our eyes on everyone, especially as we continue to innovate in this area specifically ourselves. VMware and Nicira are doing great and important things (as we work on the same problems as well), and it was certainly not intended to relegate them to just another management category.