DNS Failover or Load Balancing… or both?
Don’t worry, you’re not alone. This is a common debate for network admins with distributed resources, like multiple servers but in different data centers. Or maybe, they’re in the same data center but lack a hardware connection.
You need a way to manage traffic flow across your resources.
You’ve most likely looked in load balancing services like DNS failover and round robin. Or maybe even software load balancing.
But there are so many options and very little time… which do you choose? Which will give you the redundancy and peace of mind you deserve… without breaking the bank or over-complicating your job?
Before we can reveal the answer, we need to understand how each of these services works.
Failover is most basic of DNS-based load balancing tools, but it’s great if you need something simple and cheap. Failover automates record updates based on server availability. All you have to do is specify backup IP’s or hostnames, and failover will do the rest. It will even “failback” when your primary is back online.
The secret is integrated monitoring. We use a network of monitoring nodes to check the health of your primary endpoint every 2–4 minutes. You can read more about how DNS Made Easy managed failover updates in this blog.
It’s a great service if you need to specify a backup in case your primary server is unresponsive.
For example, if you have two servers in the same data center, but no hardware connection between them.
You can also use failover for full site failover between data centers! Or failover between different CDN or cloud providers to bypass outages.
Failover only supports one active resource at a time. So if you have more than one server in your configuration and both are your “primary” servers, you need another solution, like round robin.
You also have to consider TTL’s. Even though DNS Made Easy automatically updates all of our nameservers with the new IP for your record, resolvers will not notice the change until the TTL expires.
I like to think of round-robin like a rotor. It cycles through the different endpoints in your configuration. Meaning, each endpoint in a round robin config will be returned as often as the rest.
You can have multiple active endpoints!
Round robin is faster than software load balancers because it operates at the DNS level. Software load balancers are external and require an additional lookup, which could increase latency.
Since this is the most basic of load balancing techniques, it’s pretty limited in customizability.
Say you only want to allow a very small percentage of your users to be routed to that server? Like a roll-out, or maybe you’re phasing out old hardware. You would need a more advanced form of round robin, called Weighted Round Robin. Currently, DNS Made Easy does not offer this service, but our new product Constellix does!
Round robin lacks integrated monitoring, so if an endpoint is unhealthy or down, it will continue to be returned… meaning some of your traffic would see your site or application as down.
BUT! Constellix does have integrated monitoring with ALL load balancing services, including basic round robin.
Hopefully, this article has helped you figure out which service is best for you, but if you’re still stuck please let us know! We can most likely help you come up with a custom solution with one of our products.
Originally published at DNS Made Easy Blog.