Networking Hardware in a Virtualized World

Back on April 29 2011, I wrote one of my first thought pieces on networking; at the bottom of that post I wrote the following:

Here is another quote “Until about two years ago, we were talking about how to do identity management internally,” said Gartner analyst Gregg Kreizman. “Now, it’s about how do we get our arms around the SaaS [software-as-a-service] problem? Or we used to manage the applications but now they’re in the cloud”…so it’s leading to a never-before-asked question, “How about if we have our identities there?”

If I go back and look at the five points I highlighted on April 29, 2011, I suspect that all of these become the real important driving trends for networking companies.  I think it is going to be less about who has the fastest processor and highest throughput.  Platform choices will be made based on virtualization of many of the core functions of network device.  That is what I mean by modular software.  There is a hardware platform, but it is appliance and some appliances are bigger and faster than others.  How the networking software reaches into the virtualized I/O, interacts with the compute element and how it is controlled is what will matter.  Kind of sounds like a smartphone.  The most advance hardware device did not win the smartphone war.  The device that won the smartphone wars had the best user defined controls, the best modular software (i.e. app store) and the best analytics all of which can be found in the iPhone.  If the best hardware device had won, we would all be carrying a Nokia device.

Selling network devices is going to be about selling user definable controls and modular software (i.e. app support).  That is what end-users and service providers are going to want.  Let the users set application flow controls and SLAs.  Turn up the bandwidth and you get a bigger bill.  Security, policy and backend analytics all fall into this category.  These areas of the network are not going to be solved by selling more hardware and that is a big problem for legacy vendors.  The compute point will be extended into the network and traditional networking HW vendors will not like that trend.  Three posts in a row I will repeat what I wrote from last week:

“#2 Infrastructure Vendors Create 4th Leg: Shared cache concept and I think this is big, but this is the part of the network evolution that I wrote is really two networks: a network for humans and a network for machines to maintain a shared cache.  I believe in point #2, but maybe in a different way.  I really think that more VMs on a blade and I/O virtualization are a big way to achieve statistical gain.  I also think this is going to put pressure on the network element to do something different.  Network vendors that can integrate the network element into the I/O of the compute element are going to be very valuable.  Application delivery controllers (ADCs) become a virtual (i.e. software control) capability that is stretched across the compute/IO/network element.  This will allow it to scale and achieve maximize stat gain.  The networking vendor that figures this out with #6 below = big time winners.”

I am traveling the next couple days…no new posts until Friday.

/wrk

* It is all about the network stupid, because it is all about compute. *

** Comments are always welcome in the comments section or in private.  Just hover over the Gravatar image and click for email. **