6.9.3 Scalability measures

What this heading covers

Scalability is a shape, not a point. The measure needs at least two load levels and a property that is expected to hold, such as response time or error rate. Linear growth, sub-linear growth and collapse are different results, and a single figure cannot tell them apart. The measure should state the axis, the range and the property, and it should be taken in an environment that resembles production closely enough for the shape to transfer.

Record position: family 33 of 40 in clause 6, under 6.9 Flexibility, which carries 4 families. It is one of the 40 families a programme has to map its measures to, and its elements are listed below. The heading is 6.9.3 in the public contents; the text behind it is not public.

How a measure for this family is built

Report the property at each step of the load or data range, and the step at which it crossed the threshold. A ratio of achieved scale to required scale is defensible when both are defined at the same property. Keep the intermediate results, because the shape is the measure.

What to put in place before publication

1. Use at least three load levels and keep each result.
2. State the property that must hold as scale grows.

Evidence to keep

Two questions to ask

• What is the load or data axis and the step size?
• Which property had to hold as the load grew?

Where teams usually fall short

• A single load test at twice the current traffic, called scalable.

What the number cannot tell you

It cannot tell you where the next bottleneck is. The curve shows the shape; the architecture explains it.

What happens next

Installability and replaceability are the other two flexibility families, and both are cheaper to evidence than scalability.