Software51 mOctober 26, 2026, 14:07

Databricks Auto CDF shifts CDC cost from writes to reads

Databricks made automatic change data feed generally available on September 1, 2026, computing changes at read time from row tracking metadata instead of writing them on every commit. A 10M-row test showed 8% more data written per MERGE and 4.73 s to read one commit's changes versus 0.21 s with legacy CDF.

DEV Community

Open source

Databricks Auto CDF shifts CDC cost from writes to reads

Databricks claims Auto CDF makes MERGE and UPDATE about 15% faster
Row tracking wrote 8% more data: 547.6 MB vs 505.8 MB per MERGE
Reading one commit's changes took 4.73 s vs 0.21 s with legacy CDF
Spark 4.2 added a SQL CHANGES clause, but it needs a connector on Delta 4.4.1

#Databricks#Spark#DeltaLake

PostgreSQL 19 adds WAIT FOR LSN for read-your-writes on replicas

Security

GhostAction shows workflow write access is secret read access

Security

Linux KVM ARM64 Flaw Lets Guests Read and Write Host Memory

Business

Databricks Auto CDF shifts CDC cost from writes to reads

PostgreSQL 19 adds WAIT FOR LSN for read-your-writes on replicas

GhostAction shows workflow write access is secret read access

Linux KVM ARM64 Flaw Lets Guests Read and Write Host Memory

German auto industry in crisis: VW writes down €10B, Mercedes warns of plant closures