Read more: |
Matti Grönroos
Setting Targets for Availability
The availability targets shall have a special focus when planning for service quality according to the fit-for-purpose strategy. There are 720 hours in a 30-day month. The target availability of 99.999 per cent allows 25 seconds of downtime. The key question is: Does a 30-second service break ruin your business? Really?
Then: 99.999 per cent of what? Measuring the server CPU uptime gives rather different figures from what the end-use sees.
The availability targets are of great importance to the business both from the point of view of business continuity and financials. That is why goal setting should never be left to IT/TS only. There shall be a strong connection to the company business case. While setting the targets, a realistic approach usually gives a better result than crying for the Moon.
Before starting, each organization shall be ready to answer to two fundamental questions:
- What is the lowest acceptable availability target, possibly per priority class?
- What is the highest acceptable cost?
We remind of the general rule of thumb for risk management: Do not spend more than the estimated cost of the damage multiplied by with the probability of damage. In plain English: It is not worth investing a million euros to avoid a loss of a hundred thousand euros.
The cost to benefit curve seldom is linear but behaves according to the diagram below. It compares six different scenarios: availability achieved and their cost. When business objectives are set for the axes, only options C and D are seen to satisfy the constraints.
Based on experience, it seems that companies rarely have a precise view of what would be the right target level for critical systems availability from the business point of view.
The most common answer is: 100 per cent, of course. Since it is not an option, the correct answer can be bracketed with, for example, the following question:
What is your loss of revenue, if your systems are down for
- a second
- a minute
- an hour
- a day
- a week
- a month
The right answer usually is not the downtime divided by the length of a year multiplied by the annual revenue. If the production capacity is not fully in use, a short delay seldom causes a complete loss of revenue for the downtime. Be realistic: Especially B2B customers do not usually cancel their orders because of a short delay.
Warning: Avoid entering the one-size-fits-all thinking. Only a small part of the IT systems is usually so business critical that even a short downtime causes difficulties. It is not worth looking for the technically best solution for everything.