Internet Performance Delivered right to your inbox

DNS-A-Mania: DNS Load Balancing Vs. Hardware Load Balancing

Asking most IT professionals if they would like a low cost, high-performance load balancer that will never have to be replaced due to ancient hardware is kind of like asking Charlie Sheen if he is winning…of course!

The problem is that for years, load balancers have been strictly hardware-based and as such, cost and wear always took their physical toll. In the case of DNS-based load balancers, these problems don’t exist or are removed from the end user.

Taking advantage of the DNS layer for load balancing offers the consumer the ability to load balance anything that is issued a DNS name (which these days is virtually any core system), whether it be web servers, SMTP servers, SNMP servers, etc. This allows for round-robin load balancing based on the DNS records returned for the name query.

A smart DNS system will allocate subsequent DNS requests to different servers set up to handle the overall load, a side benefit being innate failover protection through automatic removal of non-responsive servers from the round robin queue.

Let’s get into the comparisons: DNS load balancing vs. hardware balancing. Who ya got?

Using a few more advanced features, the end user can even select exactly how much traffic they want going to specific servers or dictate that traffic coming from certain regions of the world go to specific resources (otherwise know as global server load balancing or GSLB).

While both systems achieve load balancing because of the layers they are done at, the method in which they achieve the load balancing in presents some basic differences:

  • DNS load balancing occurs at a higher level then the hardware load balancers, making them different in a few ways. A DNS load balancer alone would marshal out traffic to several different IP addresses whereas a hardware load balancer would take a single IP and split the traffic going to it to multiple machines.
  • Both a DNS load balancer and a hardware load balancer would health check on reachability and response via methods like ping, HTTP/HTTPS or SNMP but at the DNS level, it would have to verify this from the perspective of the DNS load balancer. This sits at a global level (more like a user) as opposed to the hardware load balancer which sits in the same room and is most often directly connected (and thus giving a far more specific failure location).
  • From a cost perspective, the cost of a DNS load balancer is significantly less on day one and you only pay as you go versus the amortized large upfront cost of a hardware load balancer on which you will be tied to a specific piece of hardware and its functional life as well as its lifetime usefulness. Just ask those whose vendors don’t support IPv6 on the model they own.
  • Of course, there are some benefits each can add due to their level. DNS provides things like systemic protection, global traffic management and cloud server management whereas hardware provides options like SSL offloading and web acceleration.

What it comes down to is that based on what each technology does well, there are certain times a hardware load balancer is the answer and certain times that a DNS load balancer is the answer.

The bottom line: what situations favor DNS load balancing vs. hardware load balancing?

 

  • For local load balancing (load balancing in one’s own datacenter), a hardware load balancer is really the choice. What a local hardware load balancer will do is give almost instantaneous failover for catastrophic server crashes and disk failures on site. They can also provide these devices with a measure of alarming and security.
  • For load balancing in multiple geographic regions, DNS load balancing is really the best solution. This will allow you to manage the traffic to geographic regions and handle systemic incidents through a global failover scenario (datacenter power issues or problems that occurs upstream) outside of the datacenter where the center itself would never see it but the user would.

In an end-to-end protection scenario, the ideal “best practice” would indeed be a combination of the two.

In each of your local datacenters, you would employ a hardware load balancer to deal with local system integrity and to manage the load amongst systems on premises. Then each of your data enters or cloud servers (which most likely have hardware load balancers running in the provider’s datacenters) would be load balanced geographically and protected from upstream and full system failure by a global enterprise level DNS load balancer.

I feel obliged to mention that Dynect offers the best enterprise DNS load balancer available today, but it’s the truth.

What type of load balancing you choose is specific to your business and your system topology. In the end, matching what you are trying to achieve with what benefits are derived from DNS load balancing or each hardware load balancing manufacturer will determine the correct solution for you.

Want to know more and assess what is best for you? Let’s have a discussion (sales@dyn.com) and assess whether our DNS load balancing service can offer something that would benefit you. I kinda think it would.


Share Now

Whois: Kevin Gray

Kevin Gray is a employee at Oracle Dyn Global Business Unit, a pioneer in managed DNS and a leader in cloud-based infrastructure that connects users with digital content and experiences across a global internet.