spark.sql.autoBroadcastJoinThreshold controls the estimated table-size limit under which Spark can automatically select a broadcast join. Increasing it to a value above the lookup table’s estimated size allows Spark to copy that small relation to each executor. The large fact table can then be processed locally without being fully shuffled across the cluster, often reducing network transfer and join duration. The threshold should remain within executor-memory capacity because broadcasting an excessively large relation can cause memory pressure. spark.default.parallelism controls default task parallelism, while spark.sql.shuffle.partitions controls the number of partitions used by shuffle operations. spark.executor.memory changes executor capacity but does not instruct the optimizer to select a broadcast join. Therefore, option A is the relevant configuration.

================