You are tuning a Spark job in Azure Synapse Analytics that processes large Parquet files. The job currently takes too long due to data skew. Which three actions can improve performance? (Choose three.)