Future Considerations: Software Defined
If Software Defined Networking (SDN) becomes the open ubiquitous technology that I think it will, everything changes.
That sounds dramatic, but I believe that SDN will change many aspects of how we deploy and manage networks. It also creates a completely new paradigm for security enforcement and an opportunity to think differently.
I think it will be amazing for people, for the industry, and for everything we try to do in security. It will power an Internet of Things (IoT) and forever elevate the value of data anytime, anywhere. I see SDN as the next critical step that no one will ever know happened.
When is this amazing change supposed to happen, you ask? It's already started and it will be ongoing for many years to come. It's not something where you can just flick a switch and suddenly it's all there and running; there's still lots of work to do.
But we can flick the switch ahead of time when thinking about how to build SDN strategy, and ultimately a secure one. To do this, you have to drop all current expectations of the technologies that you're running today and think about what SDN is meant to change at all levels.
To get in the right state of mind for this exercise, consider a situation where you've been running a library for many years. It's stacked full of books, magnificent collections for anyone to access and read via a book tracking system that you've spent millions on, essentially putting the Dewey Decimal System online.
Then tragedy strikes one night and the entire collection, along with the building, burns. The insurance money comes in and we are left with a real question. Does it make any sense to rebuild a building full of books, knowing what we already know about technology? Is there still a place for this? Before, due to a long history of value, this option was assumed, but when presented with, or in fact forced to recreate the library, does the design and deployment of a building of books make any sense?
I ask this question because you have to go into SDN with just that frame of mind. Ask yourself if what you're doing today makes any sense in this new design, then go a step further. Ask yourself what you need to do to empower SDN instead of looking at it from the perspective of how it might work based on how you do things today.
What It Takes
Let's flick that switch now and consider how SDN is evolving the network by walking through an SDN-enabled infrastructure from network to application.
SDN extracts network intelligence directly from switches into a centralized controller. This controller contains all the objects in the environment, from switches to applications, and everything between. The controller can send commands like “put, get, forward, delete, etc.”, as well as take in data about the state of any forwarding tables (and that's without getting into the technical details, which is another blog unto itself).
Consider a network where you can make forwarding decisions based on far more than IP data. I'm talking about simply knowing where the connection needs to be and forwarding it across any infrastructure to any application, against any security controls that you may need. Maybe you rewrite the IP header as it moves across physical connections, but that's not even necessary to consider when working with SDN as the process is abstracted away from us.
Think about what you could do with the power to forward packets based on a myriad of possible scenarios from network to application, and being able to track and protect that flow on demand. Running out of CPU and memory in one datacenter? Send the flow over to another. That one getting tapped out? Push it out into a cloud infrastructure.
New version going online of your application? No problem, as the next flow will be directed to the virtual machine running the new code. Problem with the new code? OK then, the next flow goes back to the previous version service, all on demand and orchestrated. I can't wait to see the creative things that people do with this level of program-ability and control.
The Security Perspective
How is security affected by all of this?
For starters, it's simply abstracted to a service with policy eventually moving into orchestration of that service. Don't get me wrong, security policy management remains relevant, but it moves from a dictated security policy to a monitored security policy, just not right away. And over time, traditional enterprise security policies will become less relevant. To show you what I mean, we can jump ahead to the concept of a monitored policy as part of this exercise.
Let's say that an application request comes in the form of a network call to a Web service to return data for a custom application, perhaps a new wearable armband health application. The network then checks its table to see where to send the connection, tags it accordingly, and forwards it on. In turn, the controller knows an application request is on its way for this particular service, and most likely already has a server up and running, ready to service it.
Since the controller knows how many clients it can service per virtual machine, with defined CPU and memory, it keeps spinning up new virtual machines and redirecting traffic accordingly. To include security in this process becomes a simple task. There's no need to deploy hardware and create choke points as security simply becomes another application to the abstracted network.
For example, we can forward the data based on any decision, not just the network setup, and offload a copy for traffic validation; essentially run an on-demand security scan on the same flow and let the controller know if there's a problem. Based on the orchestration decisions, the controller can have the traffic flow quarantined, blocked, redirected or just plain dropped; how and why will be tied to the value and risk of the service.
This is the point where we move from security policy management to security policy monitoring. As applications are defined and brought online, information will be collected on what data is handled by which users and corresponding threat scanning can scale up or down accordingly. It will be this on-demand delivery of security services that will enable rapid scaling of new applications.
While excited about all these the possibilities, I'm fearful of the potential nose dive that could occur if vendors try to create some form of lock in. SDN as a technology can't be stopped by this and will emerge no matter what, it's just a matter of how long it takes. Being realistic, it's just going to take a few generations of equipment to get there.
However, if we truly enable SDN from networks, along with security, and into the application, many of our current challenges go away. Not to say we won't have new issues to consider, but I'll save that discussion for another time.
Join The Conversation
We encourage you to share your thoughts, and we look forward to reading your comments. We invite you to subscribe to our blog to keep up with the latest posts of our new series.