For cash-strapped businesses, especially startups, provisioning infrastructure is often planned in accordance with perceived immediate needs. That is, servers and other infrastructure are invested in only insofar as they enable current business plans to be put into action.
In the early stages of a business, developing a reputation for reliability is key, especially in the cases where that business offers services online. Customers and clients want to be able to access the services that they pay for when they have the need. They do not want to wait and they certainly don’t want to build their own businesses around services that threaten their own reliability and customer relationships.
We can think of reliability as the ability to offer consistent levels of service over the long term. A reliable online service should not suffer downtimes and slowdowns. It’s especially crucial that they shouldn’t suffer a degradation of service at key growth periods, where demands on infrastructure may grow beyond the originally planned-for levels. It’s exactly those moments when a business should be able to capitalize on increased exposure. Bad service is a wasted opportunity for business growth.
Redundancy is simply having more capacity in the right places when a business needs it. Whether that’s extra web servers to pick up the load when a site experiences a spike in popularity, or the ability to swap in new equipment when existing infrastructure fails, having more capacity than a business needs in their day-to-day operations helps avoid the inevitable moment when something goes wrong from harming business continuity.
Lack Of Redundancy Leads To Unreliability
Let’s have a look at a couple of examples of how not to organize a business’ technological infrastructure for reliability.
Big Iron, No Fallback
We can think of this as the “putting all your eggs in one basket” syndrome. While it’s entirely possible to spend money to buy one or two servers with the capacity to handle all of the load a site or application might need, concentrating everything in those machines is very risky indeed. Sure, you have excess capacity, those super high-spec and way-over-capacity servers are not going to be strained in the near term, but, they are likely to fail at some point, and if your entire business relies on the continued functioning of a small number of machines, then it’s constantly at risk of those machines going awry.
Single points of failure are never a good idea.
The alternative to very high powered servers is to use a cluster of less expensive systems behind a load balancer. This is usually a better way to go, as there’s no single point of failure (assuming you have redundant load balancers), But, over time, as the business scales and loads increase, businesses need to stay in front of the wave of new demand by adding extra nodes on top of those they absolutely require.
Just because the stressed out servers are functioning perfectly well today, it doesn’t mean they will tomorrow. When one of the nodes in a maxed out cluster goes down, for whatever reason, the load gets transferred to the remaining nodes, which are already functioning at their maximum capacity, so performance over the entire cluster will be seriously degraded, and reliability and consistency will disappear.
Hope For The Best; Plan For The Worst
Businesses should always plan pessimistically. There’s a recognized fallacy in which people tend to plan for the best possible scenario. They imagine things going the way they want them to go, in the spirit of positivity, and plan accordingly.
Planning that way can lead to disaster. Instead, businesses should strive to think positively, but plan pessimistically to ensure that they can maintain consistency when they are hit by that unforeseen black swan.
Photo credits: Beraldo Leal