HealthOmics Nextflow Gets Resource Fallback, So Accelerator Shortages Don't Stop Your Runs!
Hi, it's Shii-chan! Today's news makes bioinformatics workflows less likely to get stuck!
AWS What's New
aws.amazon.com
What was announced?
AWS What's New announced the resource fallback directive for tasks in AWS HealthOmics Nextflow workflows. You can define an ordered list of preferred accelerator types, including an option to fall back to CPU instances at the end.
AWS HealthOmics is a HIPAA-eligible, fully managed bioinformatics workflow service for healthcare and life sciences customers.
The story so far
When your preferred accelerator was unavailable, researchers had to diagnose the problem and resubmit the run. The source says the goal is to reduce that time and keep production workflows running.
What changes
When your preferred accelerator is unavailable, HealthOmics automatically moves through the alternatives you specified, with no resubmission. Researchers and bioinformaticians can spend less time diagnosing and resubmitting runs affected by accelerator constraints!
Dive Deep
Here are the key behaviors.
- Each accelerator profile has a configurable timeout
- Shorter timeouts move quickly through the fallback order
- Longer timeouts increase the probability of reserving your preferred accelerator
- Adding a CPU profile as the final fallback increases the likelihood of your task reserving an instance
It is available in all AWS HealthOmics Regions.
- US East (N. Virginia, Ohio)
- US West (Oregon)
- Europe (Frankfurt, Ireland, London)
- Israel (Tel Aviv)
- Asia Pacific (Seoul, Singapore, Tokyo)
For configuration details, see the advanced resource configuration and Nextflow directive documentation in the HealthOmics developer guide.
Wrap-up
- A resource fallback directive lets Nextflow tasks specify accelerator priorities
- If a preferred accelerator is unavailable, the run moves to the next option without resubmission
- Per-profile timeouts control how long to wait
- A CPU profile can be the final fallback
This is for anyone running genomics pipelines on GPUs or other accelerators!
