Pipeline Runtime Calculator
Calculate total CI/CD pipeline runtime by summing sequential stage durations and taking the max across parallel stage groups.
Inputs
Comma-separated durations (seconds) for stages that run one after another, e.g. checkout, build, deploy.
Comma-separated durations (seconds) for stages that run concurrently as one parallel group; the group's duration is the slowest stage.
Time spent waiting for an available executor before the pipeline starts.
Total Pipeline Runtime
390seconds
Total Pipeline Runtime
6.50minutes
Sequential Stage Time
225seconds
Parallel Group Time
150seconds
Time Saved by Parallelism
210seconds
Step by step
Sum of sequential stages
60 + 90 + 45 + 30
= 225s
Slowest parallel stage (bounds the group)
max(120, 90, 150)
= 150s
Pipeline duration: sequential + parallel group
225 + 150
= 375s
Total runtime including queue delay
375 + 15
= 390s
How it works
Total CI/CD pipeline runtime is the sum of every stage that must run sequentially, plus the duration of the slowest stage within any parallel group (since parallel stages complete together, bounded by whichever one takes longest): total = sum(sequential_stages) + max(parallel_group_stages). Queue delay waiting for an available executor adds on top of the pipeline's own execution time.
Formula
totalRuntime = sum(sequentialStages) + max(parallelGroupStages) + queueDelay
- S_i
- Duration of sequential stage i in seconds
- P_j
- Duration of parallel stage j in seconds
- Q
- Queue delay in seconds
- T
- Total pipeline runtime in seconds
Frequently Asked Questions
Why is the parallel group's time just the maximum, not the sum?
Stages running in parallel execute concurrently on separate executors, so the group as a whole only takes as long as its slowest member — the faster stages simply finish early and wait.
How can I use this to decide what to parallelize?
Moving a slow, independent stage (e.g. a lengthy test suite) into its own parallel group alongside other independent stages reduces total pipeline time by removing it from the sequential sum, at the cost of using another executor concurrently.
Does queue delay vary a lot in practice?
Yes — on shared CI infrastructure during peak hours, queue delay can exceed the pipeline's own execution time; provisioning more executors or using ephemeral cloud runners reduces this variability.
What if I have multiple parallel groups in one pipeline?
Compute each parallel group's max independently, then add all of them plus the sequential stage sum for the full pipeline duration.