Execution classes¶
Learn how to run with different execution classes with your tool of choice.
- CLI: the Nextmv Command Line Interface (CLI).
- Python SDK: a Software Development Kit (SDK) for Python developers.
- Cloud API: an HTTP API for developers to integrate with the Nextmv platform.
Execution classes define the characteristics of how instances are executed, such as:
- cores
- memory
- maximum run time
- GPU acceleration
In a nutshell, an execution class is the type of machine that is used to run remotely. Here is the default execution class that is widely available:
| Class | Cores | Memory | Maximum run time | Premium |
|---|---|---|---|---|
6c9500mb870s |
6 | 9500 MB | 14.5 minutes | ➖ |
If an execution class is not specified, this default (6c9500mb870s) one is
used. This execution class is appropriate for low-latency, real time jobs that
require responses in under a second, up to jobs running 14.5 minutes. It scales
horizontally and does not queue requests.
Premium execution classes¶
Info
Premium execution classes must be activated on your account. Please contact us for more information.
If you want to run for longer, and with more CPU and memory available, you can use premium execution classes. These are available upon request and are subject to additional charges.
These are the premium execution classes available:
| Class | Cores | Memory | Maximum run time | Premium | GPU |
|---|---|---|---|---|---|
8c16gb12h |
8 | 16 GB | 12 hours | ✅ | |
16c60gb12h |
16 | 60 GB | 12 hours | ✅ | |
nvidia-L4-4c14gb24h |
4 | 14 GB | 24 hours | ✅ | 1x NVIDIA L4 |
You should expect your runtime to increase with this type of resource by 30-60 seconds compared with the default execution.
GPU execution classes¶
GPU-accelerated execution classes have attached NVIDIA GPUs. The
nvidia-L4-4c14gb24h class provides one NVIDIA L4 GPU per run, on an amd64
instance. Python apps that run on it should set arch:
amd64 in their manifest. To use the GPU, deploy your app with
a GPU runtime, such as
ghcr.io/nextmv-io/runtime/python-cuda:3.11-13.0. GPU runtimes can only run on
GPU execution classes. Runs that use a GPU runtime on a class without a GPU are
rejected.