The start of a New Year serves as a great reminder to re-evaluate objectives, initiatives, and technologies. As we move forward into 2019, it is important to leave the past in the past and begin building towards the future. And appliance load balancers are something enterprises need to say “goodbye” to ASAP — and for good reason!
Appliance architecture can not keep up with your business.
This is self-evident when you compare how hardware and virtual appliances compare to modern load balancing solutions. First, why do we have load balancers in the first place?
- High Availability — Applications are important to your business and they need load balancers to ensure availability. Internal and external users rely on these applications. As the number of applications and users increase, the importance of your load balancer also increases.
- Efficiency — The goal of the load balancer is to increase the efficiency of resources. Often referred to as a “traffic cop”, the load balancer routes traffic for better infrastructure utilization and better experiences for your users.
- Scale — As your applications scale, the load balancer is supposed to ensure effective and consistent functionality and security at low and peak usage.
Load balancing appliances can not provide high availability, efficiency, and scale.
Appliance Load Balancers: A Dangerous Balancing Act
Load balancing appliances are deployed in pairs. The load balancer that delivers the application is called “active” and the backup load balancer is called “standby” — you know, in case something goes wrong (e.g. power failure, hard drive failure, network failure). A typical load balancing pair looks like this:
The active load balancer usually runs between 10-15% utilization and the standby just sits there like a highly-paid backup quarterback. Nothing about this model is efficient.
When traffic to the load balancer increases it poses a threat to the availability of the application.
If the load balancer reaches capacity, it’ll topple over. Enterprises don’t have much choice but to acquire more active-standby pairs (which is inefficient) or cross their fingers and hope traffic doesn’t exceed capacity.
When we look at this model in the data center with physical appliances or in the cloud with virtual appliances, it shows what a mess appliance load balancers create.
Most business units are siloed and are unable to share resources. If a load balancer pair needs additional capacity, they are unable to use resources from other infrastructure. And if they were, most application and load balancer configurations are snowflake deployments. It is non-trivial to re-deploy an application with consistent policies and configurations in a different environment. This is why appliance load balancers require “fleets” of load balancing pairs for each silo.
Load balancing appliances can not provide high availability, efficiency, and scale.
What Must Change About Load Balancer Appliances?
Modern enterprises need modern load balancers. And that means that we need to re-imagine how a load balancer should work. We need to move from appliance architecture to a software-defined architecture. For example, an appliance load balancer manages the control plane and the data plane discretely from each appliance. A software-defined solution creates an elastic load balancing fabric that separates the control plane and data plane, which provides better availability, efficiency, and scale for your applications.
How Does a Software-Defined Load Balancer Work?
First, and most noticeable, no more active-standby. Because this load balancing architecture is not based on appliances, there is no such things as a “standby” load balancer. You have infinite scale so you can scale up or down based on need, which is the most efficient way to ensure availability.
This model also boasts self-healing and resiliency. That means, that if a load balancer has an issue, another can be spun up immediately to replace it.
Second, the separate control plane simplifies load balancer management as it can centrally control all the load balancers for any application or business unit, across any cloud or data center. Now you can manage load balancing for all your applications without increasing complexity. The controller can give each application and business unit the functionality it needs without requiring discrete load balancer fleets, even if the applications are deployed in public clouds like AWS or Azure.
Load balancing today still has the same requirements: High Availability, Efficiency, and Scale. But appliance-based load balancers can no longer live up to the task. It’s time to say “goodbye” to the load balancer appliance.
