Does 99.5% planned uptime = 99.5% actual uptime?

If you are evaluating cloud financial management applications, a common question is “how often does the system go down?” For most of the leading cloud vendors, they have a strikingly similar answer: 0.5% of the time, phrased as “99.5% uptime”. That sounds very impressive, but does that really mean your application will be up and running 99.5% of the time? The short answer is no.

Why is the answer no? Simply because:

A)     Many vendors define the equation to get the outcome they desire.

B)     The math is tricky!

Here are three very important questions you need to ask the vendors to truly understand your potential downtime and evaluate the different products:

1) How do you define planned downtime?

There are 744 hours in a 31-day month.   Without digging further into how the vendor calculates the uptime percentage, it would be natural to do simple math and quickly determine that .005 downtime equals four hours per month. This seems very reasonable, on the surface, but remember math is tricky and the vendor controls the math.   Some vendors calculate their uptime percentage as:

Actual Hours System Up divided by (Hours in the Month minus Planned Downtime)

The key question to ask is “how much planned downtime do you have in any given month?”   You will find a wide array of answers to this question. For example, one leading vendor plans for 40 hours of planned downtime a month to apply patches, fixes, and general system maintenance.   The very reasonable four hours of allowed downtime in their marketing equation equals 44 hours, or almost two days, of actual downtime a month. Very tricky!

2) What are the exceptions to planned downtime?

So now that we know how the software vendors works the math to create a great marketing statistic, another really important questions to ask is:

Are there exceptions to your planned downtime equation?

One would think the answer to that question would be no, as it is a service level agreement and the vendor needs to meet the planned uptime (even with their funny math!). Surprise: many of these vendors have exceptions. For example, one vendor states that if they give you notice in advance that the system is going to be down then that downtime does not count towards the planned downtime percentage.   The 44 hours of monthly downtime in the scenario above could really be any downtime number.

3) How do major releases impact planned downtime?

Another exception might be for major upgrades or releases.   Vendors often exclude this downtime from the equation.   Major upgrades could take days and may occur up to four times a year. You should ask the vendor if the upgrades are an exception to the planned downtime calculation, and if so, how long is the downtime?   You should also ask for a historical trend of their downtime for upgrades and their future plans to reduce this number of hours.

And you also need to take into account difference between on-premises, hosted, single-tenant cloud and multi-tenant cloud.   A multi-tenant cloud vendor must apply updates to one and only one application.   The other three types of software vendors need to apply patches and fixes either to all of their environments individually or in groups. This takes time and equals more downtime.  

Conclusion:

Software vendors can easily meet the planned uptime percentage that is proudly displayed in their marketing slides, but 99.5% does not necessarily mean 99.5%.   As an informed buyer of these systems, you should ask the key questions discussed in this article to determine what the actual downtime is for the system, and ask for statistics that show system up and down time irrespective of planned and exception downtime.   By doing a little middle school math, you can level the playing field across software vendors.