Arthur Cole spoke with Paul Andersen, senior marketing manager, Array Networks.
To many observers, application delivery networking (ADN) and software-defined networking (SDN) appear to be two sides of the same coin. Although they appear to focus primarily on improving application performance across distributed architectures, they actually operate in very different ways, according to Array Networks' Paul Andersen. This doesn't mean they both can't be effective tools in cloud-based dynamic data environments. But it does mean they will need to be implemented properly in order to function in a complementary manner.
Cole: There's been a fair amount of discussion lately surrounding application delivery networking (ADN) and software-defined networking (SDN). What are the primary differences between these two architectures, and will they necessarily be mutually exclusive as networks and data infrastructure evolve?
Andersen: To my mind, the main difference between ADN and SDN is one function vs. approach. Application delivery networking started out with simple load balancing and has since evolved to provide integrated traffic management, application acceleration and security features. Whereas load balancing, caching, compression, SSL acceleration, WAN optimization, SSL VPN, Web application firewalls and many other functions were distinct at one point in time, today they are collapsing into a single solution called ADN. What’s more, where these products were largely appliance-based in years past, today they are also available as software solutions for virtual and cloud environments. The purpose of ADN is to ensure that applications behave the way they were designed to, as if the user, device and application were all in close proximity. In today’s world of mobility and cloud computing, users and apps are virtually never in close proximity and are often changing locations. ADN assures performance, security and availability in this dynamic application environment.
Software-defined networking on the other hand is less of a set of functions in and of themselves, but rather an approach to routing data in a manner that is better suited to the mobile and cloud dynamics mentioned above. By separating out the control plane, it is possible to create policies and logic that are far more capable of maximizing the use of resources in a VM-driven data center. It is also possible to get a handle on customers/users/end points that rapidly change their locations on the network and the underlying infrastructure. The way I see it, they are not mutually exclusive at all; ADN and SDN appear to be more cross-cutting in nature and complementary within private and public cloud environments. Both will integrate with Open Flow or cloud management systems to spin up and spin down resources as needed to support application performance and customer/organizational requirements and to ensure the most efficient way of connecting users and applications within a highly dynamic environment.
Cole: What changes can we expect in the application delivery controller, particularly as enterprises seek to extend dynamic data environments onto the cloud?
Andersen: The availability of application delivery controllers in software/virtualized form factors, capable of running with sufficient performance and features on a standardized VM infrastructure. Management extensibility will be essential to integrating with cloud management systems and SDN systems for service automation. For enterprises modernizing their data centers or building their own private clouds, this integration could take the shape of tighter integration with VMware vCenter or similar management systems from the provider of the virtual infrastructure. For cloud providers, this integration is more likely to occur with their own cloud management system, which may or may not take advantage of OpenFlow-like technologies.
Cole: Ultimately, though, we're talking about the logical decoupling of applications and data from underlying infrastructure. Will we ever get to the point at which applications will be able to compile their own network resources as needs arise?
Andersen: I do think the lines between ADN and SDN have the potential to become blurred, depending on adoption and how things unfold. Going back to load balancing, the core/original function of an ADC was to direct traffic. If ADCs were to develop integration with SDN, load balancing could be preempted by a more advanced and dynamic instruction for directing traffic. In addition, it is feasible that certain routing instructions will spawn the needed underlying networking functions, etc. Or that there could be increasing synergy between applications, cloud management, SDN and ADN to work in concert in a dynamic fashion to ensure optimal use of network resource and maximum application performance, security and availability.