Quick test, not a rigorous benchmark — just logging numbers before I forget them. Tagged
lab-note — shorter and rougher than the full write-ups, filterable via the tag search on
the posts page.
Setup
- Uploading a 500MB file from a home connection in Jakarta
- Target bucket in
ap-southeast-1(Singapore) - Compared: direct
aws s3 cpvs the same command against the*.s3-accelerate.amazonaws.comendpoint - 5 runs each, averaged, network otherwise idle
Results
| Method | Avg time | Avg throughput |
|---|---|---|
| Direct upload | 71s | ~7.0 MB/s |
| Transfer Acceleration | 68s | ~7.4 MB/s |
Takeaway
Basically a wash for this route. Makes sense — Jakarta to Singapore is already short and
well-peered, so there’s not much latency for Transfer Acceleration’s edge-routing to save.
I’d expect a bigger gap uploading to a bucket in us-east-1 or eu-west-1 from here, where
the request would otherwise cross more of the public internet before reaching AWS’s backbone.
Next: rerun this against a us-east-1 bucket to see if the gap opens up. Also want to test
with a batch of small files instead of one large one — Transfer Acceleration’s benefit profile
is supposed to differ there.