100% Uptime Guarantee – Realistic or Just a Fable?

First off, let me just answer the question here - Real 100% uptime? The answer is no. The first question you are probably asking is: why is that? “I see companies like Amazon offering nine 9’s of uptime” or other companies promising me 100% uptime.

Before I answer that question, let’s take a look at a couple of news making events that had a significant impact on the customers that they served.

“Milions of Hotmail users cut off by Microsoft cloud failure, millions of users of Hotmail and other Microsoft services worldwide were unable to access their online accounts this morning after the firms cloud suffered a major technical failover.” Source from Telegraph

“Google on Friday blamed a recent Google Docs outage on a real-time collaboration update that exposed a glitch in its system...Unfortunately this change exposed a memory management bug which was only evident under heavy usage.” Source from PC Mag

Most people consider Microsoft and Google to be the big players in the cloud war, and competition and service levels are fiercely coveted to ensure their uptime is better then the other. But what do these recent outages tell us? No matter how much planning there is, or the amazing skills of the people building these systems, you have to consider that for regular consumers who use services like Gmail, Hotmail, Yahoo Mail or other free cloud-based services--it's free. You get what you pay for. However, these services are integral to these companies’ brands and represent products that maintain a high stickiness factor with end-users. As an example, how many times have you looked at your free email account today? My answer is 36 times. If my mail goes down, it is annoying but honestly, it isn’t the end of the world.

But let’s change equation and ask what happens when you are a business that relies on paid products that utilize the cloud? Even further, what happens when you are a billion dollar business running on the cloud? Apple’s recent announcement that iCloud will run on Microsoft Azure and Amazon Web Services should make everyone excited and nervous at the same time. Because if there is an outage and you lose access to your files, or your music, you at least can feel better knowing that Apple is suffering along with you as this article describes.

Studies continue to show that cloud services routinely offer availability above 99%, which is higher than that usually provided by in-house IT departments, and for companies where a percentage point can make a difference of a million dollars, this is significant. Cloud providers today are continually working to eliminate all single points of failure and prevent outages, but there still needs to be a backup plan. As a hosting company, we are continually looking for redundancy across our entire service offering - redundant switches, redundant power, redundant bandwidth, redundant network cards, redundant servers - you get the idea. We are continually working to minimize single points of failover, particularly for our server and SaaS customers.

So what’s a good SLA (service level agreement)? The obvious answer is a number that matches your business requirements and does not negatively impact your business. There is a reason we offer our web hosting customers different service levels. For some organizations, 99.9% is more then satisfactory. For other clients of ours, they’ve asked for 99.999%+ uptime and know that any type of outages will have negative consequences that surpass the cost of hosting in that type of environment.

It is important to understand that with each nine you add to your service level, you are significantly increasing your costs to minimize outages for your customers and most studies have shown that five nines is the tipping point for increased costs. At that point, you are adding new layers of redundancy to your service offering. As an example, in building our new VMware cloud hosting service, we made the business decision to target five nines as the desired service level for our customers and build our solution with enterprise grade hardware from Dell and the leading virtualization platform from VMware. We built redundant fault tolerant server clusters that gives our customers better availability and failover in the event of an outage, as well as decreased downtime and smaller time to recovery. However, in doing so we had to invest more than 3x’s as much money into the service offering to guarantee that amount of uptime. Consider this, five nines is about five minutes of unscheduled downtime a year.

To give you an idea, there are 525,600 minutes in a year and 4 minutes a year represents .00000076 of that total.

So is the cloud reliable and safe? Yes. Is it more reliable then classic hosting? In a lot of cases, yes. Similar to what we always tell our customers, you do need to be prepared. Having a plan B always makes sense, and no matter how much preparation goes into it you can never cover all of the contingencies.

“Lightning Causes Amazon, Microsoft Cloud Outages....we understand at this point that a lightning strike hit a transform from a utility provider to one of our Availability Zones in Dublin, sparking an explosion and fire. Normally, upon dropping the utility power provides by the transfomer, electrical load would be picked up by backup generators. The transient electric deviation...disabled the backup generator plan.“ Source from CRN News