Perhaps not, but then you don't really need two arms, either.

Anyone who has broken an arm, especially the one with their dominant hand, can tell you that while it probably wasn't any fun they learned to manage just fine with a single arm until the other one healed. They'd also likely tell you that they were glad they had two arms in the first place, because without two they'd have been hard-pressed to do anything productive with only one arm and that one broken, in a cast, unusable.

There's a lot of FUD out there regarding whether or not you need two load balancers. After all, the chances that the primary load balancer is going to fail is minimal and the secondary one more often than not ends up simply consuming resources while waiting in the wings. Some people go ahead and buy into redundancy and then deploy in an active-active scenario in order to get the most out of their investment, despite the potential problems that could arise from such a deployment should one of the load balancers fail. Some people don't deploy a redundant pair, figuring that the risk of failure is minimal enough that if it does happen, they'll simply replace it and move on.

So do you really need a redundant pair of load balancers?

The answer to that reallly depends on your level of comfort as well as service level agreements and the importance of the application/site you're load balancing. It's a matter of risk management; of determining whether or not you can afford - financially or otherwise - a failure.

  • If you could lose money due to downtime, then you need a redundant pair; the cost of a second load balancer is minimal when compared to the potential loss of revenue should a single load balancer fail. This is not much different than paying for insurance; sure, you probably won't experience a major catastrophe that will rack up massive hospital and doctor bills, but if you do it's nice to know that you've got it covered.
  • If you have service agreements with partners or a distribution channel then you need a redundant pair of load balancers. The cost to your reputation and impact on your business depends on meeting those service level agreements. A secondary load balancer ensures that even if the primary fails that you can still meet those agreements.
  • If your business relies on the applications being load balanced, you need a redundant pair. Consider the marked drop in productivity that would occur if you deployed a single load balancer in front of the applications your employees use every day to conduct business and that load balancer failed. How would your employees get any work done? How would they interact and help customers? How would it impact your supply chain? A second load balancer provides reassurance that in the event of a failure on the part of the primary that your business will keep running. That keeps employees working, customers happy, and business processes running smothly.
  • If you're load balancing just a few servers and the applications/site on those servers isn't mission critical, isn't going to cost you money if they aren't accessible, and if downtime isn't going to destroy your reputation you can probably get by without a second load balancer. I'm not sure this type of environment really exists - except in my basement - but if so, you're okay with just one.

Deploying a redundant pair of load balancers is important to providing assurance of availability. It's true that a single load balancer - especially if it's from a reputable vendor - is unlikely to fail. It's also unlikely you'll break your arm, but if you do you'll be glad you have two of them.

Imbibing: Coffee