# Computation Checkpointing & Result Storage

> Available S3 endpoint: `a-s3.aur.lu` (France, Albi).

## Ray (checkpoints)

Distributed compute framework for Python/AI workloads. Store training checkpoints and artifacts on S3‑compatible storage for resilience and portability.

- Homepage: [Ray](https://www.ray.io/)
- Docs: [Ray Docs](https://docs.ray.io/)
```python
import ray

ray.init()
uri = "s3://my-bucket/ray-checkpoints"
options = {
  "aws_access_key_id": "<YOUR_ACCESS_KEY>",
  "aws_secret_access_key": "<YOUR_SECRET_KEY>",
  "endpoint_url": "https://a-s3.aur.lu",
}
# Use uri/options in your trainers or custom checkpoint logic
```

## Snakemake (remote S3)

Workflow engine for reproducible science. The S3 remote provider enables cloud storage of intermediate and final outputs.

- Homepage: [Snakemake](https://snakemake.github.io/)
- Docs: [Snakemake Docs](https://snakemake.readthedocs.io/)
```python title="Snakefile"
from snakemake.remote.S3 import RemoteProvider as S3RemoteProvider
S3 = S3RemoteProvider(host="a-s3.aur.lu", secure=True)

rule all:
    input: S3.remote("s3://my-bucket/pipeline/output.txt")
```

## Prefect 2.x

Orchestration platform for data/ML with blocks for S3‑backed flow and result storage.

- Homepage: [Prefect](https://www.prefect.io/)
- Docs: [Prefect AWS](https://docs.prefect.io/integrations/prefect-aws/)
```python
from prefect_aws.s3 import S3Bucket

block = S3Bucket(
    bucket_name="prefect-results",
    endpoint_url="https://a-s3.aur.lu",
    aws_access_key_id="<YOUR_ACCESS_KEY>",
    aws_secret_access_key="<YOUR_SECRET_KEY>",
)
block.save("aurora-s3")
```