ARA-C01 Question 280
Single answerSnowpipe StreamingA financial services company is ingesting high-volume trade events from a custom Java application into Snowflake. The application requires sub-second data availability for downstream dashboards, must avoid managing staged files, and should minimize warehouse usage for ingestion. Architects are evaluating Snowpipe Streaming. Which approach best meets these requirements?
- A
Use the Snowpipe Streaming high-performance architecture with the Snowflake Ingest SDK to stream rows directly into Snowflake-managed ingest infrastructure, then query the target table as data becomes available.
- B
Use classic Snowpipe with an internal stage and configure auto-ingest so files are loaded continuously without requiring a virtual warehouse for ingestion.
- C
Use a task to run COPY INTO every few seconds from a temporary stage populated by the Java application, because tasks provide the same low-latency ingestion model as Snowpipe Streaming.
- D
Use the Kafka connector in file-based mode even though the source is a custom Java application, because Snowpipe Streaming requires Kafka as the transport layer.
Show answer and explanation
Correct answer: A
Explanation
The key architectural distinction is that Snowpipe Streaming is intended for low-latency streaming ingestion directly from applications or supported connectors, without first writing files to an internal or external stage. This makes it well suited for event-driven workloads such as trade events where dashboards need fresh data quickly. In contrast, classic Snowpipe is fundamentally file-oriented: it continuously loads newly arrived files from stages and is excellent for micro-batch file ingestion, but it does not eliminate staged file handling.
For SnowPro Advanced: Architect candidates, the decision point is matching workload characteristics to the right ingestion pattern:
- Choose Snowpipe Streaming when producers need row-level or very low-latency ingestion and want to avoid stage/file orchestration.
- Choose classic Snowpipe when upstream systems naturally produce files and event-driven file loading is sufficient.
- Avoid replacing streaming ingestion with scheduled COPY INTO tasks when low latency and operational simplicity are required, because that adds warehouse dependency and stage management.
This aligns with Snowflake best practices and product documentation describing Snowpipe Streaming as a low-latency streaming ingestion mechanism using the Snowflake Ingest SDK or supported connectors, while classic Snowpipe remains the managed continuous file-loading service.
- A. Correct.
Correct. Snowpipe Streaming is designed for low-latency row-based ingestion directly from client applications without the need to land files in a stage. Using the Snowflake Ingest SDK from a custom Java application aligns with the requirement to avoid file management. Snowpipe Streaming also uses Snowflake-managed compute for ingestion rather than requiring a customer-managed virtual warehouse for the ingest operation itself, which helps minimize warehouse usage for loading. This option best matches the stated need for sub-second availability and direct integration from a custom app.
- B. Incorrect.
Incorrect. Classic Snowpipe does support continuous file loading from stages and does not require a virtual warehouse for the pipe ingestion itself, but it is still file-based. The scenario explicitly requires avoiding staged file management and achieving the lowest possible latency from a custom application. Classic Snowpipe is not the best fit when the producer can stream rows directly and wants near-real-time ingestion without micro-batching files.
- C. Incorrect.
Incorrect. A task running COPY INTO from staged files introduces both file management overhead and additional latency. Tasks also require compute resources to execute SQL on a schedule, and this pattern is not equivalent to Snowpipe Streaming's row-based ingestion path. Even aggressive scheduling would generally be a workaround rather than the recommended architecture for sub-second streaming ingestion.
- D. Incorrect.
Incorrect. Snowpipe Streaming does not require Kafka. While Snowflake provides streaming support for certain connector patterns, a custom Java producer can use the Snowflake Ingest SDK directly. Choosing Kafka solely to enable Snowpipe Streaming is a misconception and adds unnecessary infrastructure when the source application can integrate directly.