-
Notifications
You must be signed in to change notification settings - Fork 241
How to improve the performance of tpcds with spark3.3.3? #408
Hello!I build the environment of spark3.3.3 and blaze2.0.8, then i do some tests based on 100G tpcds data,however,I did not receive any benefits compared to not using blaze,this is my spark-sql command:
spark-sql --master spark://xxxx:xxxx --conf spark.sql.extensions=org.apache.spark.sql.blaze.BlazeSparkSessionExtension --conf spark.shuffle.manager=org.apache.spark.sql.execution.blaze.shuffle.BlazeShuffleManager --conf spark.blaze.enable.smjInequalityJoin=true
I want to know if I need to add any other parameters. Looking forward to your reply!
All reactions
Replies: 1 comment
If you wish to assess the acceleration effects of Blaze, here are some details that are easily overlooked:
-
Review the data format of your TPC-DS tables. Ultimately, only Parquet and ORC formats are optimized for performance acceleration, with Parquet currently offering better support. If you're not experiencing significant benefits with Blaze, I suggest examining the data format for potential improvements.
-
It is advisable to establish a standard test environment. The test environment can significantly impact performance, and unresolved issues can adversely affect overall revenue.
-
Adjusting Parameter Settings. You may refer to the following guidelines: Run TPC-DS benchmark with Spark/Blaze