Have you ever wondered why it seems Google, Facebook, Netflix, Amazon, or another large website almost never seems to go down? As cloud services become more pervasive, “downtime” is becoming a thing of the past.
Downtime, as the word implies, is an indication that a website has crashed and isn’t available online. Website owners, obviously, don’t like downtime.
In “The Different Types of Web Hosting, and Why it Matters,” I explain the pros and cons to web hosting options. But none of them are immune to failure.
If a web hosting company offers a “99% monthly uptime guarantee,” that implies that the company guarantees a website will be available online for 99% of the time. There are 720 hours in a 30-day month. 99% of 720 hours is a little less than 713 hours. That means the company guarantees your website won’t be down for more than 7 hours.
(Develop CENTS offers a 99.9% uptime guarantee to all of our shared hosting clients).
But no downtime is obviously more preferable than even a little (That is an extremely rare and sometimes impossible feat. I am reminded of a tweet I read recently that said “If your uptime is 100%, your monitoring is broken).
Enter High Availability Web Hosting.
High Availability (HA) Web Hosting is much more expensive than regular hosting, because it requires two web servers, as well as at least one “load balancer” which is typically a 3rd server that “balances” website traffic between the two web servers.
Although nothing is immune to downtime (including High Availability options), HA reduces that risk substantially. With the right setup, if 1 server becomes overwhelmed by traffic or otherwise becomes unavailable, the other server(s) will pick up the slack.
Huge companies like Google, Amazon and Facebook even go beyond traditional HA deployments, and use a combination of DNS, routing, and load balancers to balance website traffic “on the fly” and ensure the closest and most appropriate web server is serving a particular person in a particular place on earth.
Anything can go wrong, and no company is immune to human mistakes and outside attacks that could potentially bring down a server or network, but HA deployments are a good step in the direction of creating an “always available” infrastructure.