manideep4597 opened a new issue, #17524:
URL: https://github.com/apache/iceberg/issues/17524
### Feature Request / Improvement
### Feature Request / Improvement
#### Summary
`GCSFileIO` / `GCSOutputStream` always uploads via GCS `WriteChannel`
(`storage.writer(...)`), which uses **resumable uploads**. For small objects
this adds unnecessary round-trips and latency compared to a single-shot /
simple upload (`storage.create(...)`).
`S3FileIO` already switches between single `PutObject` and multipart based
on size (`s3.multipart.part-size-bytes`, `s3.multipart.threshold`). GCS has no
equivalent.
#### Motivation / Use case
Iceberg writes many relatively small objects (manifests, metadata JSON,
delete files, etc.). For these, resumable-upload setup cost often dominates
transfer time.
Google Cloud Storage Java guidance also recommends:
- small content → `storage.create(blobInfo, bytes)` (single request)
- large / streaming content → `storage.writer(...)` (resumable)
Today Iceberg always takes the resumable path, which is suboptimal for small
files and creates a feature gap vs S3.
#### Current behavior
In `GCSOutputStream`, uploads always open a write channel:
```java
WriteChannel channel =
storage.writer(
BlobInfo.newBuilder(blobId).build(), writeOptions.toArray(new
BlobWriteOption[0]));
gcpProperties.channelWriteChunkSize().ifPresent(channel::setChunkSize);
stream = Channels.newOutputStream(channel);
### Query engine
_No response_
### Willingness to contribute
- [ ] I can contribute this improvement/feature independently
- [x] I would be willing to contribute this improvement/feature with
guidance from the Iceberg community
- [ ] I cannot contribute this improvement/feature at this time
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
