Send2Bucket vs Amazon Data Firehose
Amazon Data Firehose is AWS's managed delivery stream: you put records with the AWS SDK and it buffers them into S3, an Iceberg table or S3 Tables in your account. Send2Bucket does the same delivery from a plain HTTPS or OTLP endpoint, and to buckets and catalogs outside AWS too.
Short answer: if every producer already has AWS credentials and every destination is in AWS, Firehose is cheaper per GB. Choose Send2Bucket when producers send OpenTelemetry or plain HTTP with an API key, or when data has to land outside AWS.
At a glance
| Send2Bucket | Amazon Data Firehose | |
|---|---|---|
| How you send | POST /v1/events with an API key; OTLP over HTTP or gRPC | PutRecord / PutRecordBatch with AWS SigV4 |
| OpenTelemetry | Native OTLP endpoint | No OTLP endpoint; run a Collector in front |
| Buckets | S3, Google Cloud Storage, S3-compatible (R2, MinIO, B2) | S3 |
| Tables | Iceberg REST catalogs, S3 Tables, Supabase Analytics Buckets, DuckLake | Iceberg through AWS Glue, S3 Tables, plus warehouses such as Redshift and Snowflake |
| Object format | gzip NDJSON with a per-record envelope | Raw records, or Parquet and ORC with format conversion |
| Where it runs | Hosted by Send2Bucket; writes into your account | A stream you create in each AWS account |
| Price | Plans and billing | $0.029/GB ingested (first 500 TB, 5 KB billing increments); Iceberg delivery $0.045/GB; Parquet conversion $0.018/GB |
When Firehose fits better
- All producers run in AWS and already sign requests with IAM.
- You need Parquet or ORC objects rather than NDJSON.
- Price per GB decides it: at 1 TB a month Firehose costs roughly $29 to $47 before storage.
When Send2Bucket fits better
- OpenTelemetry without a Collector. Point an OTLP exporter at Send2Bucket; Firehose needs a Collector or agent to translate OTLP into
PutRecordcalls. - Clients outside AWS. Browsers, edge functions and partner systems send with one API key instead of AWS credentials.
- More than one cloud. The same stream can deliver to Google Cloud Storage, an S3-compatible bucket or a non-AWS Iceberg catalog.
- One acknowledgement rule. A
200means the batch is stored durably and will be delivered, with per-batch delivery IDs in every record. See Delivery semantics.
Moving from Firehose
Records arrive as gzip NDJSON under s2b/stream=…/date=…/hour=…/, each wrapped in an envelope with batch and delivery IDs. See Delivery format. Athena reads them with partition projection, and S3 Tables destinations appear in the AWS Glue Data Catalog as before.