HealthOmics runs your private workflow tasks in an omics instance. HealthOmics provides a variety of instance types to
accommodate different types of tasks. Each instance type has a fixed memory and vCPU configuration (and fixed GPU
configuration for accelerated computing instance types). The cost of using an omics instance varies depending on
the instance type. For details, see the HealthOmics Pricing
For tasks in a workflow, you specify the required memory and vCPUs in the workflow definition file. When a
workflow task runs, HealthOmics allocates the smallest omics instance that accommodates the requested memory and vCPUs.
For example, if a task needs 64 GiB of memory and 8 vCPUs, HealthOmics selects omics.r.2xlarge
.
We recommend that you review the instance types and set your requested vCPUs and memory size to match the instance that best meets your needs. The task container uses the number of vCPUs and the memory size that you specify in your workflow definition file, even if the instance type has additional vCPUs and memory.
The following list contains additional information about vCPU and memory allocation:
Container resource allocations are hard limits. If a task runs out of memory or attempts to use additional vCPUs , the task generates an error log and exits.
-
If you don’t specify any compute or memory requirements, HealthOmics selects omics.c.large and defaults to a configuration with 1 vCPU and 1 GiB of memory.
-
The minimum configuration that you can request is 1 vCPU and 1 GiB of memory.
-
If you specify vCPUs, memory, or GPUs that exceeds the supported instance types, HealthOmics throws an error message and the workflow fails validations
-
If you specify fractional units, HealthOmics rounds up to the nearest integer.
HealthOmics reserves a small amount of memory (5%) for management and logging agents, so the full memory allocation might not always be available to the application in the task.
-
HealthOmics matches instance types to fit the compute and memory requirements that you specify, and may use a mix of hardware generations. For this reason, there can be some minor variances in task run times for the same task.
These topics provide details about the instance types that HealthOmics supports.
Topics
Standard instance types
For standard instance types, the configurations aim for a balance of compute power and memory.
HealthOmics supports the 32xlarge and 48xlarge instances in these regions: US West (Oregon) and US East (N. Virginia).
Instance | Number of vCPUs | Memory |
---|---|---|
omics.m.large | 2 | 8 GiB |
omics.m.xlarge | 4 | 16 GiB |
omics.m.2xlarge | 8 | 32 GiB |
omics.m.4xlarge | 16 | 64 GiB |
omics.m.8xlarge | 32 | 128 GiB |
omics.m.12xlarge | 48 | 192 GiB |
omics.m.16xlarge | 64 | 256 GiB |
omics.m.24xlarge | 96 | 384 GiB |
omics.m.32xlarge | 128 | 512 GiB |
omics.m.48xlarge | 192 | 768 GiB |
Compute-optimized instances
For compute-optimized instance types, the configurations have more compute power and less memory.
HealthOmics supports the 32xlarge and 48xlarge instances in these regions: US West (Oregon) and US East (N. Virginia).
Instance | Number of vCPUs | Memory |
---|---|---|
omics.c.large | 2 | 4 GiB |
omics.c.xlarge | 4 | 8 GiB |
omics.c.2xlarge | 8 | 16 GiB |
omics.c.4xlarge | 16 | 32 GiB |
omics.c.8xlarge | 32 | 64 GiB |
omics.c.12xlarge | 48 | 96 GiB |
omics.c.16xlarge | 64 | 128 GiB |
omics.c.24xlarge | 96 | 192 GiB |
omics.c.32xlarge | 128 | 256 GiB |
omics.c.48xlarge | 192 | 384 GiB |
Memory-optimized instances
For memory-optimized instance types, the configurations have less compute power and more memory.
HealthOmics supports the 32xlarge and 48xlarge instances in these regions: US West (Oregon) and US East (N. Virginia).
Instance | Number of vCPUs | Memory |
---|---|---|
omics.r.large | 2 | 16 GiB |
omics.r.xlarge | 4 | 32 GiB |
omics.r.2xlarge | 8 | 64 GiB |
omics.r.4xlarge | 16 | 128 GiB |
omics.r.8xlarge | 32 | 256 GiB |
omics.r.12xlarge | 48 | 384 GiB |
omics.r.16xlarge | 64 | 512 GiB |
omics.r.24xlarge | 96 | 768 GiB |
omics.r.32xlarge | 128 | 1024 GiB |
omics.r.48xlarge | 192 | 1536 GiB |
Accelerated-computing instances
You can optionally specify GPU resources for each task in a workflow, so that HealthOmics allocates an accelerated-computing instance for the task. For information on how to specify the GPU information in the workflow definition file, see Task accelerators in a HealthOmics workflow definition.
If you specify a GPU that supports multiple instance types, HealthOmics selects the instance type based on availability. If both instance types are available, HealthOmics gives preference to the lower cost instance.
G4 instances aren't supported in the Israel (Tel Aviv) Region. G5 instances aren't support in the Asia Pacific (Singapore) Region.
G6 and G6e instance types
HealthOmics supports the following G6 accelerated-computing instance configurations. All omics.g6 instances use Nvidia L4 or Nvidia L4 A10G GPUs.
HealthOmics supports the G6 and G6e instances in these regions: US West (Oregon) and US East (N. Virginia).
Instance | Number of vCPUs | Memory | Number of GPUs | GPU memory |
---|---|---|---|---|
omics.g6.xlarge | 4 | 16 GiB | 1 | 48 GiB |
omics.g6.2xlarge | 8 | 32 GiB | 1 | 48 GiB |
omics.g6.4xlarge | 16 | 64 GiB | 1 | 48 GiB |
omics.g6.8xlarge | 32 | 128 GiB | 1 | 48 GiB |
omics.g6.12xlarge | 48 | 192 GiB | 4 | 192 GiB |
omics.g6.16xlarge | 64 | 256 GiB | 1 | 48 GiB |
omics.g6.24xlarge | 96 | 192 GiB | 4 | 192 GiB |
All omics.g6e instances use Nvidia L40s GPUs.
Instance | Number of vCPUs | Memory | Number of GPUs | GPU memory |
---|---|---|---|---|
omics.g6e.xlarge | 4 | 32 GiB | 1 | 24 GiB |
omics.g6e.2xlarge | 8 | 64 GiB | 1 | 24 GiB |
omics.g6e.4xlarge | 16 | 128 GiB | 1 | 24 GiB |
omics.g6e.8xlarge | 32 | 256 GiB | 1 | 24 GiB |
omics.g6e.12xlarge | 48 | 384 GiB | 4 | 96 GiB |
omics.g6e.16xlarge | 64 | 512 GiB | 1 | 96 GiB |
omics.g6e.24xlarge | 96 | 768 GiB | 4 | 96 GiB |
G4 and G5 instances
HealthOmics supports the following G4 and G5 accelerated-computing instance configurations.
All omics.g5 instances use Nvidia L4 A10G, Nvidia Tesla A10G, or Nvidia Tesla T4 A10G GPUs.
Instance | Number of vCPUs | Memory | Number of GPUs | GPU memory |
---|---|---|---|---|
omics.g5.xlarge | 4 | 16 GiB | 1 | 24 GiB |
omics.g5.2xlarge | 8 | 32 GiB | 1 | 24 GiB |
omics.g5.4xlarge | 16 | 64 GiB | 1 | 24 GiB |
omics.g5.8xlarge | 32 | 128 GiB | 1 | 24 GiB |
omics.g5.12xlarge | 48 | 192 GiB | 4 | 96 GiB |
omics.g5.16xlarge | 64 | 256 GiB | 1 | 24 GiB |
omics.g5.24xlarge | 96 | 384 GiB | 4 | 96 GiB |
All omics.g4dn instances use Nvidia Tesla T4 or Nvidia Tesla T4 A10G GPUs.
Instance | Number of vCPUs | Memory | Number of GPUs | GPU memory |
---|---|---|---|---|
omics.g4dn.xlarge | 4 | 16 GiB | 1 | 16 GiB |
omics.g4dn.2xlarge | 8 | 32 GiB | 1 | 16 GiB |
omics.g4dn.4xlarge | 16 | 64 GiB | 1 | 16 GiB |
omics.g4dn.8xlarge | 32 | 128 GiB | 1 | 16 GiB |
omics.g4dn.12xlarge | 48 | 192 GiB | 4 | 64 GiB |
omics.g4dn.16xlarge | 64 | 256 GiB | 1 | 24 GiB |