By Patrick MeLampy
Prediction #1: The Downfall of Virtual Networks
My first prediction is that the term “virtual” will no longer be sexy in 2017. Virtual is old, middle aged, and getting a bit thick in the middle. It just seems that for years, anything declared virtual was perceived to be better. Virtual has been defined to mean “not physically existing as such, but made by software to appear to do so.” To be sure, virtualized compute markedly lowered the operational cost of individual operating systems and servers. It also led to automated management of applications – another benefit that has saved customers big bucks. Virtual storage was a natural follow-on, and we have seen the cost of storage go down while the quality and performance of storage go up.
As we enter 2017, “virtual networking” – a trend that’s been around for a few years, is now poised to take off. However, the step-function cost and agility improvements associated with virtual compute and storage may not fully grace the concept of virtual networking.
Here’s why: Virtual applications that use virtual storage are enabled by REAL (physical) networking, which provides the command and control infrastructure as well. So after we use the REAL network to launch a virtual application that attaches to virtual storage over a REAL network, we then want to create a virtual network for the application to communicate on. The virtual network will receive its command and control over the REAL network. How is this simpler than before?
No worries, automation will save us. Just stir in some orchestration to allocate network addresses and virtual LANs, provision port groups on virtual switches, throw in some multi-cast to create virtual broadcast services. And don’t forget to provision the virtual access control lists and virtual firewalls!
To make all of this easier we have created service function chains that will allow us to define a sequence of virtual network functions to be processed in order. This will permit insertion of a virtual load balancer, or a virtual DPI device, or a virtual security device between an application server and the REAL internet.
The good news is that there are many additional tools and capabilities being added to help debug and understand the virtual networking layers in real time, including virtual wire shark, virtual port monitors, etc.
“Virtualization” became sexy because it changed the game. From what I see, “virtual networks” don’t change the game – they make it more cumbersome, which is hardly sexy.
Prediction #2: IT Seeks a New Network Paradigm
My second prediction is that by the end of 2017, the IT industry will finally understand why AWS, Facebook, and Google have all replaced their networking software completely – rather than adopt layers of network virtualization. These companies have built their own proprietary solutions because no commercial models worked. They saw that the true disruption was re-defining the network stack in a way that made sense for their business – not furthering the existing paradigm.
Compute, storage, and application users are all connected by REAL networks. The chance that all three will be under the same network command and control is near zero. Virtual networking that crosses REAL network borders doesn’t exist and simply isn’t going to work. If the industry wants to do what AWS, Facebook, and Google are doing, then the industry must act, and act soon. The first key step in transforming our networks is to admit we have a problem.
Patrick MeLampy is the Co-Founder and Chief Operating Officer at 128 Technology.
The original article can be found here.