The uptime guarantee explained
28 August 2018
When browsing for an IT service, it’s common to see in the SLA a 99% uptime guarantee. Occasionally you might spot a 99.9% uptime guarantee. And rarely you might even find a 99.9999% uptime guarantee, but it’s typically a sales ploy. Whilst these numbers sound good, what do they actually mean for your business? As it turns out a 99% guarantee just isn’t good enough anymore. Here’s why…
If your business uses a service with a 99% uptime guarantee that means you should expect:
- 14 minutes, 24 seconds of downtime every day;
- 1 hour, 40 minutes and 48 seconds of downtime every week;
- 6 hours, 43 minutes and 12 seconds of downtime every month and
- 3 days, 15 hours, 21 minutes and 36 seconds of downtime every year
It’s important to note that these numbers rely on downtime only ever occurring during opening hours. In the ‘real world’ about 1/3 of downtime would occur during work-hours (in the actual real world it’s more than 1/3 because higher loads on services at this time increase the chance of a service outage.)
But even if we take the more ‘realistic’ 1/3 value you still end up with 1 day, 5 hours, 7 minutes and 12 seconds of unplanned downtime a year. This would cost 98% of businesses over £2,233,000 in lost revenue according to data from an ITIC survey. And this is without mentioning the costs down the line if clients decide to leave based on the poor performance.
For many companies, this level of downtime is unacceptable. Not only for the financial reasons but for the impact it has on their image. Many customers expect a responsive service 24/7. So even a short period of downtime can permanently taint a user’s opinion of a company meaning that better guarantees are necessary.
99.9% uptime guarantees (referred to as three-nines) have become the new standard for most digital services. They provide decent availability with only a small amount of unplanned downtime. With a ‘three-nines’ plan you should expect the following:
- 1 minute, 26.4 seconds of downtime every day;
- 10 minutes, 4.8 seconds of downtime every week;
- 40 minutes, 19.2 seconds of downtime every month and
- 8 hours, 44 minutes and 9.6 seconds of downtime every year
It might be surprising to see the impact of an additional 0.9%. But the reason that the change is so drastic is that 90% of the 99% guarantee’s downtime is removed with the addition of another nine. Because each nine reduces downtime by 90%, uptime guarantees become exponentially more effective.
A ‘four-nines’ guarantee gives even better rates of uptime. If you use a service offering a 99.99% guarantee you should expect:
- 8.64 seconds of downtime every day;
- 1 minute, 0.48 seconds of downtime every week;
- 4 minutes, 1.92 seconds of downtime every month and
- 52 minutes, 24.96 seconds of downtime every year
Less than an hour of downtime a year sounds good (and it is good). But there’s the big factor of cost to remember. As strange as it sounds, at a certain point it becomes more cost effective to allow downtime to occur because the cost of a more reliable service outstrips the losses caused by an outage.
There’s still the reputational damage to consider though so, as it is with everything, it’s a matter of balancing things correctly.
- 0.864 seconds of downtime every day;
- 6.048 seconds of downtime every week;
- 24.192 seconds of downtime every month and
- 5 minutes, 14.496 seconds of downtime every year
‘Five-nines’ is currently regarded as the gold standard for uptime terms because of how small the margin for error is. A service running with exactly 99.999% uptime would take 11.45 years before the service would have been down for an hour.
Unfortunately, it’s statistically impossible to have guaranteed 100% uptime because something can always go wrong. But despite this you still sometimes see a service offering this. Quite simply whenever you see a 100% guarantee it is either going to be a sales ploy, an overconfident service provider or a scam.
In a lot of cases, a 100% uptime guarantee is either backed with nothing of substance or nothing at all in their service level agreement (SLA). This means that if they don’t reach the promised 100% uptime (which they won’t) you get nothing as compensation.
An example of this would be if the SLA has a strict classification of what can be claimed on. An SLA could state that for it to be claimed on, a period of downtime must be longer than 2 hours and be caused by a technical fault on their hardware. This means that an hour-long outage wouldn’t be claimable. Neither would an outage caused by bad weather disrupting your connection, even where it causes damage to your business.
It’s important to always thoroughly read the terms and conditions for any service provider you sign up to. And equally important to consider alternative points besides just the uptime percent. Because factors like the SLA, hardware, security and compensation can be just as, if not more, important.
Although a 100% guarantee is impossible, we could hypothetically get pretty close in the future. A ‘nine-nines’ uptime guarantee would allow you to enjoy:
- 0.0000864 seconds of downtime every day;
- 0.0006048 seconds of downtime every week;
- 0.0024192 seconds of downtime every month and
- 0.0314496 seconds of downtime every year
You just experienced a decade of downtime with a ‘nine-nines’ uptime guarantee. In fact, if you’ve been reading this at an average speed, you’ve experienced about 4,000 years of downtime between the start of this blog and now. That’s pretty impressive.
Unfortunately, we’re a long way off getting hardware setups sophisticated enough to reliably give this level of availability. And even if we were, it would be ridiculously expensive because of the ludicrous amount of redundant hardware, backup, maintenance, monitoring and security systems, power infrastructure and planning that such a system would require.
That’s not to say it’s impossible, but it’s a long, long way from where we are right now.