Intermediate
Open
Pro
Computing Wall-Clock Time: Pipeline vs. Barrier
A workflow needs to process 4 independent modules, each through 3
stages: scan, refactor, test. The per-module, per-stage
durations (in minutes) are:
| Module | scan | refactor | test |
|---|---|---|---|
| M1 | 2 | 5 | 1 |
| M2 | 4 | 2 | 2 |
| M3 | 1 | 3 | 4 |
| M4 | 3 | 1 | 1 |
Assume unlimited parallel agent capacity (no queueing).
- Compute the total wall-clock time under a barrier design, where all 4 modules must finish a stage before any module starts the next stage. Show the per-stage waits.
- Compute the total wall-clock time under a pipeline design, where each module moves through its own stages independently.
- Explain, referencing the specific numbers, why the gap between the two grows as you add more modules or more stages, not just as a coincidence of this particular example.
Share this question