Why DNS Load Balancing is Still a Necessity

Let us begin with a story about a potential customer—a fictional European company that manufactures running gear, including shoes, socks, t-shirts, and more. It started in Spain, became popular there, and quickly expanded to other countries in Europe and the Middle East.

This company has launched numerous websites and applications to support its various regions and stores. Its IT leaders decided early on to host the sites and apps in different environments, involving multiple public cloud providers and local data centers, all so that the company’s regional teams could be autonomous.

The company now seeks to control those different estates more easily and to enable adaptative apps to run more smoothly to help with its blue-green deployments. (These are deployments where separate but identical environments are created, with one running a current app version and the other running a new version.)

Let’s see how DNS Load Balancing (DNSLB) can help this enterprise achieve its goals.

DNS is one of the few critical protocols of the Internet. Whenever a user needs to reach an application or a website, a DNS request will be performed to locate those resources. Without DNS, applications or websites are not reachable.

DNS (Domain Name System) may be a familiar term to most people, but DNS Load Balancing, also known as GSLB (Global Server Load Balancing), probably is not. In a nutshell, it is a more intelligent version of DNS, enabling the steering of traffic based on applications’ health and client source IP addresses.

Most applications today are no longer hosted in single locations but in multiple environments (e.g., on-premises and the public cloud, or per region). While this allows for better service to users, it also brings a need to route customers’ traffic to the closest data center

Anycast routing is typically used in these situations. But it requires all traffic to go the provider’s network, which forbids architectures leveraging locations with different providers.

Additionally, customers need a way to detect a failing backend—that is, to avoid sending requests to it and causing a bad user experience.

This is where DNS Load Balancing (DNSLB) comes in handy. Here’s why:

• DNSLB offers regular health checks to make sure that the endpoints participating in an application are active.
• Not only this, but the health checks can also look for specific keyworks to conclude that a backend is healthy—eliminating backends that are reachable but in a degraded way, and are not serving content properly.
• Geo-IP support means that end users can be served with the most relevant content.
  • For example, all users will be sent to an anycast web service, but users in countries far away from PoPs (points of presence) may be sent to a local data center or regional cloud provider.
  • Or, if users can't be sent outside their country or region, geo-based policy ensures that they are served content from the relevant place (this applies to the European Union's GDPR requirements as well).
• DNSLB can also be part of a disaster recovery plan, allowing the routing of users to a fallback location in case the first location fails.

In our example of the company manufacturing running gear, with its IT spread across different environments, DNSLB is a smart way to bring flexibility. That’s because it deals only with the DNS layer while the rest of the traffic flows to the chosen destination (e.g., public cloud, private data center, etc.). This enables it to avoid being tied to a specific vendor or cloud provider.

Also, thanks to the different load-balancing algorithms available, the company can deploy new versions of its apps using a blue-green approach (gradually switching users to a newer version), or even performing a canary deployment—a deployment that cuts over a small subset of servers or nodes first before including others—by using the “ratio” algorithm.

Having DNSLB as SaaS comes with distinct advantages over an on-premises solution:

• Customers don’t have to worry about provisioning any hardware or network and rack capacity.
• Distributed Denial of Service (DDoS) protection is included—which means customers don’t have to take care of it themselves.
• Anycast routing gives end users access to the DNS servers closest to their locations, therefore improving latency and providing a better user experience.

Also, of course, a REST-based API should be available for customers to easily automate and integrate with their existing tools and systems.