# Performance Tuning

Optimize throughput and latency when interacting with Aurora’s S3-compatible endpoints.

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

## Uploads

- Use multipart uploads for files >100 MB.
- Parallelize parts (4–16 workers typical; tune to network/CPU).
- Choose part sizes 8–64 MB; larger parts reduce overhead, smaller parts increase parallelism.

## Downloads

- Use ranged GETs to parallelize large object downloads when needed.
- Reuse HTTP connections (keep-alive) in SDKs.

## Listing at scale

- Use prefixes and delimiters to partition listings.
- Paginate using `ContinuationToken`.

## Example: aws CLI parallel config

```bash
# ~/.aws/config
[profile aurora]
max_concurrent_requests = 20
max_bandwidth = 50MB/s
```

## Networking

- Place compute close to the selected endpoint region.
- Prefer persistent connections and HTTP/2 where SDK supports it.