The year 2009 may be remembered as the year technologies died. First Anne Thomas Maynes of Burton Group declared SOA dead, and more recently Mark Fabbi of Gartner announced the death of load balancers. The difference in the obituaries is striking: Maynes declare an entire architectural model dead while Fabbi merely declares the death of a product, not the technological concepts behind it. Load balancers may be dead, the concept of load balancing lives on as a critical foundation for more advanced and valuable features available in the load balancer’s evolutionary replacement: the application delivery controller.

Where Maynes gives no guidance on what organizations should be replacing SOA with, Fabbi does a good job of explaining not only what, but why organizations should  be focusing on delivering applications instead of simply distributing packets via aging load balancer solutions.

The position of a load balancing solution in the data center is such that it places the solution in front of the applications for which it is providing availability. That means it takes on the role of an intermediary, a data center application broker, if you will. An intermediary can certainly provide basic load balancing duties, but it can also serve as gatekeeper, guardian, performance coach, and well-informed guide for the applications it delivers.

We are already in the throes of data center consolidation as a means by which the operational and capital expenses of IT can be reduced, resulting in a more financially efficient infrastructure. Virtualization is often used as a means by which unused compute resources can be leveraged to reduce waste as well as the physical number of machines in the data center, which subsequently has a positive effect on the costs associated with power, heating, cooling, and space.

Consolidation at the infrastructure level is also possible, but rarely considered. An integrated application delivery controller combining web application acceleration, optimization, and security with its core load balancing capabilities reduces the number of physical individual solutions necessary which reduces power, heating, cooling, and space costs as well as the overhead of management and the annual costs of maintenance and support.

Dare I mention that an integrated solution also reduces the number of physical devices through which traffic must flow, which necessarily reduces latency and thus improves application performance?

Robin Layland in “A new era in application delivery” sums up the situation well:

The problem is that implementing all these necessary but separate solutions [security, consolidation, availability, performance] will cause the number of appliances to grow, increasing cost and complexity.

What is needed is an integrated solution that can classify, prioritize and control traffic at the application layer while also providing most, if not all, of the needed application layers services. The banner this solution flies under is Application Delivery Networking.

The return on investment of an application delivery controller is much higher than is possible for load balancers simply because an application delivery controller can provide additional functionality that cannot be found in aging load balancers. As Layland goes on to say, “The key to application delivery meeting the new challenges is to integrate in a single solution the right mix of application visibility, acceleration, optimization, security and SLB.” The application delivery controller (ADC) does just that. It integrates core load balancing with acceleration, optimization, and security solutions. 

The point Fabbi appears to be making is not that load balancing is an antiquated and outdated technology, but that it’s necessary to look beyond just load balancing to application delivery in order to architect an optimized environment. Application delivery is necessary as a flexible, adaptable foundation for emerging data center architectural models like cloud computing and virtualization.

Load balancing will continue to be an important and integral function within the data center, perhaps  more so in the coming years than ever before. But like the load-bearing wall in a building it is not, on its own, a complete solution to the problems a house is intended to address. Without other walls and doors and a roof, it’s not much of a house at all. In fact, unless you live on a beach in the Caribbean I’m guessing that you wouldn’t risk your health and well-being on a house built solely on load-bearing walls.

So why would you essentially do the same to your applications? The house that load balancing built is held up by strong, proven technology – load balancing. But you really need to finish the house before your applications move in.