Avi Networks shows you don’t need special hardware to load balance

Another infrastructure aspect that proves software can do much of what hardware does

Avi shows you don’t need special hardware to load balance
Thinkstock

A trend that has grown over the past decade or so to become pretty much the default view of infrastructure is to “software-ize” functionality that was formerly the domain of specialist hardware.

It all started (arguably) with the idea of virtualization. Instead of needing a physical server for every task, software would allow numerous virtual servers to run on a piece of physical kit. The upshot of virtualization of servers was that far greater efficiencies could be generated, and utilization rates went from being dismal to almost absolute. All good outcomes if you’re worried about the economics of technology.

+ Also on Network World: Is infrastructure relevant in a software-defined world? +

But it wasn’t just compute that got this dose of software goodness. Next came storage, then networking. And seemingly the sky is the limit as to what parts of infrastructure can be made virtual. (And in the next realm of innovation, we have serverless computing where, in effect, stuff happens without even having to think about servers—physical or virtual. But that’s another story.)

Virtualized load balancing

Today's example is load balancing. As the name implies, load balancing in infrastructure describes a function much like the words mean in the physical world. A balanced load in the physical world means that all of that stuff you’re schlepping on your trailer, or all of the passenger weight on an airplane, is balanced from side to side. And in the virtual world, a load balancer is a device that distributes application traffic over a variety of servers. They’re used to optimize and increase the reliability of applications, and they are an increasingly important tool.

But here’s the thing. Load balancing can be virtualized, too, and in doing so, flexibility is increased and cost is decreased when compared to traditional appliance-based methods.

Avi Networks is one vendor attempting to reinvent what traditional infrastructure looks like. Avi Networks' platform runs on x86 servers, virtual machines or containers and offers a bunch of functionality—all aimed at creating a dynamic pool of resources and then automating those resources for the best outcomes possible.

Avi Networks today released the results of a trial that aimed to see just how scalable its elastic fabric is. In the test, Avi Networks’ software delivered 1 million SSL transactions per second (TPS) to a single web application, automatically scaling up services in just 10 minutes from zero to peak traffic, after which it scaled down to normal levels. The results would seem to debunk a common myth that companies must spend lots of money on proprietary hardware load balancers to handle large transaction volumes or high throughput.

This focus on load balancing is even more relevant given the high incidence of cyber attacks and the move to increase ever-greater proportions of internet traffic. Over 50 percent of traffic on the internet, including Netflix movies, is now encrypted. This presents new challenges for network managers and application architects who must ensure that their applications are secure and responsive for end users. The growing amount of encrypted traffic puts enormous strain on load balancers that must offload or decrypt these SSL-encrypted packets before they reach the application.

Software-defined architecture for load balancing

Avi Networks conducted the test on SSL traffic to demonstrate how a software-defined architecture for load balancing can work for the growing volume of encrypted traffic. In the test, the company ran its software in front of a single web application to load balance a continuously growing volume of SSL-encrypted transactions from over 320 clients. Avi Networks' built-in visibility and analytics dashboard was used to monitor the build-up of transactions and the application response.

According to Avi Networks, the test is relevant for a number of use cases, including:

  • Elasticity to handle seasonal changes (e.g., Black Friday) in application traffic
  • Handle DDoS attacks by elastically autoscaling until the attack is mitigated
  • The ability to scale beyond the data center by automatically bursting to the cloud
  • Eliminating over-provisioning caused by expensive, purpose-built hardware load balancers
  • Scaling up or down dynamically through integration with native cloud APIs

My POV

In load balancing, as before with compute, storage and networking, the incumbent vendors with the generally expensive, closed and inflexible proprietary hardware and software solutions are being squeezed by these newer software-defined approaches.

Avi Networks’ case study will, no doubt, be argued against by the traditional vendors who will explain why it is a specific use case and doesn’t work in other particular contexts. That argument may or may not be true, but the important point from all of this is that software is enabling an entirely new level of flexibility while optimizing the economics of technology. And that’s a compelling proposition whichever way you look at it.

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Now read: Getting grounded in IoT