A software-defined network: is it an evolution or a revolution in networking? The hype of SDN has been around for several years, but as yet it doesn’t seem to have managed to get much traction outside of the MSPs and Fortune 500 companies with regard to SDN, and telcos with regard to SD-WAN. When, if ever, will the SDN meltwater reach the fertile plains of the LME?
For this, we really need to look to history.
The original Internet was based on a research project by DARPA designed to provide a resilient interconnected network of computers in a postapocalyptic world, in order to allow a continuing military response. A grand scheme indeed. Fast-forward a few years, and think about how you—or if you’re a millennial, your parents—connected to the Internet via dial-up networking over the POTS system (plain old telephone system). Back in the day, connecting to the Internet (forget about the speed) meant tying up the phone line—yes that’s right kids, you couldn’t do both. The original Internet was an overlay network running on the telephone network, which obviously meant that the telephone network was the interlay network.
The next innovation was ISDN (Integrated Services Digital Network). This allowed the delivery of always-on Internet and fully available telephone capability. Again, this was delivered over the POTS system, with a terminator at the endpoint to split the two channels. ISDN was a digital overly over legacy last-mile analogue equipment.
This was a new overlay, and the previous overlay network became an underlay. Once again, this is a suggestion of evolution rather than revolution. To be fair, this is more akin to how an average network admin/architect looks at the world. Network teams are not ones for Big Bang revolutions.
Modern SDN vendors make a big deal about the concept of overlay networks, but these are nothing new in the world of networking. TCP/IP overlaid the old SONET networks to provide resilience. They also make a big deal about the concept of an uncoupled control plane, stating that logic is contained in the manager. TCP/IP decoupled the journey by concentrating logic in the endpoint and target devices.
SD-WAN just appears to be an overlay to MPLS, which appears to be VPNs on steroids; now we are getting to the crux of the issue for those companies that are pushing back against SDN. They truly see no value in the proposition of moving to SD-WAN over what they are currently utilizing. Their networks and intersite connections are just too small to pay the price for the new technology. They bought into virtualization because it visibly saved them money. They could physically see the benefit in the reduction of racks in their data centers and in the reduction in ancillary costs regarding cooling and power.
SDN does not have that immediate and visible benefit. It is an overlay too far. Concepts like an uncoupled control plane and automation do not really sit well with companies when talking about their networks. Agility and flexibility are not really words that resonate with network teams tasked with keeping the lights on. When you configure and forget your network, day-to-day changes are related to floor moves and the occasional new server. Your network does not significantly change over a considerable period of time.
Next is the issue that the two main players in the market, Cicso with its ACI and VMware with NSX, don’t have coherent messages. Cisco has had ACI in its arsenal since 2013, but selling it will cut off the head of the hardware fatted cow, notwithstanding that it requires upwards of a $100K investment into extra hardware. ACI is only available on the NX900 range of hardware, and it is very unlikely to trickle down into the other Nexus switches.
VMware’s Achilles’ heel is the fact that it is virtualization only and, up until recently, VMware only. VMware has reintroduced the multihypervisor version NSX-T. It now supports hypervisors based on KVM.
Other vendors, like Big Switch, are useful in bridging that chasm between physical and virtual devices when coupled with VMware NSX, but from a consumer’s point of view, this is just additional expense.
The fact is that SDN and SD-WAN are hard sells. They are, in most cases, prohibitively expensive and lacking in full data center visibility. This creates islands of SDN surrounded by traditional networking. It is safe to say that on the whole, Tier 1 providers and MSP have moved to an SDN model for their customers. To be fair, it makes sense at the scale of network changes these customers undertake. The vast majority of smaller clients just do not have a requirement for the scale and flexibility of SDN; in fact, these customers are more likely to have moved to the cloud before they move to SDN.