The underlying connector uses bigquery storage api services to pull bigquery table data at scale using spark workers.

The storage api services is cheaper and faster than the traditional Bigquery query api services enabling faster & cheaper Bigquery migration incrementally in a continuous fashion.

Pre-requisite

Need spark 4.0.0 or Databricks runtime version 15.3 & above.
pip install bigquery-spark-streaming-connector

Pyspark usage:

from streaming_connector import bq_stream_register

query=(spark.readStream.format("bigquery-streaming")
 .option("project_id", <bq_project_id>)
 .option("incremental_checkpoint_field",<table_incremental_ts_based_col>)
 .option("dataset",<bq_dataset_name>)
 .option("table",<bq_table_name>)
 .option("service_auth_json_file_name",<service_account_json_file_name>)
 .option("max_parallel_conn",<max_parallel_threads_to_pull_data>) #defaults max 1000
 .load()
 ## The above will ingest table data incrementally using the provided timestamp based field and latest value is checkpointed using offset semantics.
 ## Without the incremental input field full table ingestion is done.
 ## The service_account_json files need to be available to every spark executor workers in the '/home' folder using --files /home/<file_name>.json options or using init script (using Databricks spark)
 
 (query.writeStream.trigger(processingTime='30 seconds') 
 .option("checkpointLocation", "checkpoint_path")
 .foreachBatch(writeToTable) #your target table write function
 .start()
)