Streaming
ReleaseWorth evaluating
Real-Time Mode reaches GA for Lakeflow pipelines, with ForEach sinks to operational stores
From Databricks blog
Real-Time Mode on Spark Declarative Pipelines (Lakeflow pipelines) is generally available, giving millisecond-range latency for streaming flows. A real-time pipeline runs in continuous mode with an update flow whose trigger is RealTime, writing to a sink; a ForEach sink can write to Lakebase, Redis or another OLTP store. Binning several real-time flows into one pipeline and a native Lakebase sink are listed as next steps.
What to re-read
More from Streaming
- PreviewSeveral real-time flows can share one serverless pipeline; both new options are in Beta
- ReleaseAsynchronous processing for Python transformWithState is generally available
- ReleaseOn-demand state repartitioning is generally available for stateful streaming
- ReleaseOn-demand state repartitioning added for Structured Streaming