Constant work pattern

I recently came across a pattern that was new to me.

The name “constant work” says it all: you do the same amount of work regardless of the load or the failures the platform is experiencing.

Typically, you always do the work sized for the worst case, even when there’s almost nothing to do; this way, whatever happens, that part of the system is stable, reliable and predictable.

Another interesting property you might want when applying the pattern is anti-fragility: the work stays fixed, but its contents shrink as things fail.

Of course, it doesn’t fit every use case. The pattern only works when it’s cost-effective, so the worst-case payload should be small and fixed-size.

The best-known reference on this is Reliability, constant work, and a good cup of coffee.