Monitoring cloud costs
Monitor cloud costs to manage resources effectively and prevent unexpected expenses when running pipelines in Seqera Platform.
Resource labels
Use Resource labels in your compute environments to annotate and track the actual cloud resources consumed by a pipeline run. Resource labels are applied to the resources spawned during a run and sent to your cloud provider in key=value
format.
Seqera cost estimate
Run details include an Estimated cost display. This is the total estimated compute cost of all tasks in the pipeline run.
The Seqera cost estimator should only be used for at-a-glance heuristic purposes. For accounting and legal cost reporting, use resource labels and leverage your compute platform's native cost reporting tools.
The compute cost of a task is computed as follows:
See also: cost, start, complete, cpus, and memory in the task table.
Seqera uses a database of prices for AWS, Azure, and Google Cloud, across all instance types, regions, and zones, to fetch the VM price for each task. This database is updated periodically to reflect the most recent prices.
Prior to version 22.4.x, the cost estimate used realtime
instead of complete
and start
to measure the task runtime. The realtime
metric tends to underestimate the billable runtime because it doesn't include the time required to stage input and output files.
The estimated cost is subject to several limitations:
-
It doesn't account for the cost of storage, network, the head job, or how tasks are mapped to VMs. As a result, it tends to underestimate the true cost of a pipeline run.
-
On a resumed pipeline run, the cost of cached tasks is included in the estimated cost. This estimate is an aggregation of all compute costs associated with the run. As a result, the total cost of multiple attempts of a pipeline run tends to overestimate the actual cost, because the cost of cached tasks may be counted multiple times.
For accurate cost accounting, you should use the cost reporting tools for your cloud provider.