Snowflake makes Elastic Channels generally available

Snowflake has made Elastic Channels for Snowpipe Streaming generally available, opening a direct path from applications, devices and services into standard Snowflake tables and Snowflake-managed Apache Iceberg tables.

Snowflake says the service supports up to 20 GB/s per table and can make data queryable in as little as five seconds, depending on the workload. Those figures are vendor claims, not independently reproduced benchmarks. The company also said its internal testing reached 40 GB/s into a single table. It did not disclose the test method and said benchmark details would come in a later post.

The Elastic Channels design gives each streaming pipe one implicit channel instead of requiring producers to manage channel operations. The channel accepts concurrent appends from multiple producers, while the SDK batches records and retries transient failures. Snowflake handles ingestion scaling and the channel lifecycle, leaving applications without individual channels to create, name, open, close or recover.

Elastic Channels targets continuous streaming. Snowflake has a separate Bulk INSERT path for populated Hybrid Tables for qualifying bulk writes routed through object storage.

A successful acknowledgment means Snowflake has durably buffered an append, not that the rows are already queryable. Snowflake then validates the data, can apply transformations or pre-clustering, and commits it to the table. The service is generally available across AWS, Google Cloud and Microsoft Azure commercial regions.

The service provides at-least-once delivery without ordering guarantees. Snowflake documents that retries after an ambiguous failure can create duplicates; applications needing ordered or exactly-once recovery should use Named Channels. Producers must also retain unacknowledged events when their delivery requirements demand it because the SDK in-memory buffer does not survive a process crash. The GA API requires Snowpipe Streaming SDK 1.8.0 or later for Java, Python and Node.js.