Batch Size describes the amount of things that are handled in one step in a process.
The goal within DevOps Organizations is to work in distinctly Small Batch Sizes, that is: achieve Single Piece Flow, which means in practice to build and release each commit separately. Do not bundle commits. Because:
- Debugging: If a build fails, it’s much easier to debug the cause (and fix it) if a single commit (from a single author) is the culprit, than debugging a large amount of commits (that are not causal, that are from multiple authors) at the same time
- Failure Impact: If a build fails, and the Batch is not proceeded, the only consequence is that the (failing) commit itself is not live. With a bunch of commits together, a single mistake in any of them holds all of them back - working code is prevented from going live & bringing value (i.e. Work in Progress is growing for those that are unnecessarily waiting)
The antithesis of this idea is accumulation of code changes over a long time (weeks or even months), which are then released together, with a Big Bang.
Sources
