Task compute and resources
When you define tasks in a HealthOmics workflow, you specify the compute, memory, and container resources that each task requires. HealthOmics allocates the appropriate instance type for each task based on the resources you request.
In the workflow definition, define the following for each task:
-
CPU and memory – The number of vCPUs and amount of memory for standard, compute-optimized, or memory-optimized instances. See Task resources in a HealthOmics workflow definition.
-
Container image – The Amazon ECR container image for the task. See Container images for private workflows.
-
GPU accelerators – Optionally, a GPU accelerator type to allocate an accelerated-computing instance (G4, G5, G6, or G6e) for tasks that benefit from GPU processing. See Task accelerators in a HealthOmics workflow definition.
-
Advanced resource configuration – Optionally, an ordered list of accelerator types (GPU) to search and execute at run time. HealthOmics tries each accelerator profile in order until one succeeds, enabling fallback from one accelerator type to another or from accelerators to CPU. See Advanced resource configuration.
Note
HealthOmics matches instance types to fit the compute and memory requirements that you specify. If you don't specify any compute or memory requirements, HealthOmics defaults to 1 vCPU and 1 GiB of memory for a CPU instance.