SDN, vNET and the Future
My father-in-law loved beta. He still believes that it was a better format than VHS. And it was. But, unfortunately, based on the fact that you couldn’t record a full football game (2 hour limit) and some of the licensing costs, JVC’s VHS format won out. Lower quality, but longer run times and less expensive. If you have a box of beta tapes somewhere, you understand the value of standards and the pain of placing your bets too early. I’m sure I still have “important” data on 4mm DAT backup tapes somewhere, but accessing that data would mean reinstalling a drive into my servers and finding some software to read it, not an easy task.
While Software Defined Networking (SDN) is starting to get traction, there is a looming issue out there: who will win? But more important than who will win, is who won’t. Making the wrong call on SDN can have significant downstream costs for your business. Anyone that deployed “pre-N” or “pre-G” wireless only to find out that their access points weren’t compatible with the final standard knows this pain.
There are two main goals for SDN, the first is changing the focus of traffic management from the device to the application, and the second is delivering a more flexible and automated mechanism for making network-level changes that requires less “hands on” contact with physical devices.
To that first goal, NextIO is already delivering some of that functionality today with the vNET I/O Maestro. For the typical customer the challenge has become the abstraction between the Virtual Machines and the physical hardware. In the early days of computing, one application ran on one server which had a single pool of I/O devices and management was not a challenge – the application to I/O relationship we simple and straightforward. However virtualization changed all of this and customers soon saw that the hypervisor became the bottleneck for I/O. NextIO understood this and brought I/O virtualization to the mainstream, allowing a better manageability interface between VMs (and by virtue of their ‘1 application to 1 VM’) and the network resources. As a leader in understanding I/O and virtualization, NextIO is helping customers do a better job of managing their I/O traffic without having to change how they deploy and manage VMs today – minimizing the impact of change in the data center.
To the second goal, by pooling and sharing I/O, NextIO can already bring the ability to allow changes to happen, remotely, from a console. This means that complicated tasks like re-provisioning, moving resources, expanding resources and re-assigning resources can be done easily, in seconds, not minutes, hours or even (in some cases) days. Best of all, there is no need to physically touch the servers or switches. Provisioning for both Ethernet and Fibre Channel (two distinct networks) happens when the system is installed, and never needs to happen again, no matter how often changes are made to the servers and pooled I/O.
SDN is still evolving today and there is really only one established “open source” standard that multiple vendors are behind, and that is OpenFlow. But today, OpenFlow is being challenged by the network heavyweights like Cisco and Juniper that are trying to create their own SDN methodologies. How this battle will end is anyone’s guess at this point, which is why many people are talking about SDN, but few are actually betting on the technology with their budgets and deploying SDN. Companies would rather take a step back, let the dust clear and then deploy based on the leader, helping ensure compatibility moving forward.
NextIO is already delivering much of the functionality that customers want in SDN, but they are able to take advantage of this functionality without having to change their end of row switches or their core switches. Nor do they need to change their deployment methodology, drivers, applications, software stacks or governance models. The biggest challenge that customers see today in SDN is that adding “SDN aware” switches and routers to their network is a very costly proposition. And Cisco recently pointed out that most of the higher end (i.e. expensive) products are deployed but not fully amortized, so don’t expect people to start pulling out non-depreciated equipment to replace it with SDN equipment any time soon. The transition to SDN equipment will take a long time. After all, VMware introduced GSX and ESX back in 2001, and now, almost 12 years later, with the clear benefits of virtualization known and no longer being debated, virtualized servers have just now crossed over the 50% threshold in the data center. Enterprise technology changes happen slowly because of the cost and impact. Smaller, less disruptive technologies (like vNET’s I/O virtualization) can happen more quickly because of the limited scope of impact when deploying – everything is contained within the rack and no new fabrics are being introduced; the decision to deploy is much safer and easier.
As SDN continues to evolve, we’ll be watching the developments and taking the appropriate steps to ensure that our products can live in an SDN world – no matter what it looks like. The key for NextIO customers is that they can take advantage of the benefits of SDN today without having to worry that they are betting on the wrong horse or that they will have to make a fundamental shift down the road.
