-
Integrating Auron with Remote Shuffle Services (RSS)Auron supports integration with external Remote Shuffle Services to improve shuffle performance and scalability. Currently, the following versions are supported: Apache CelebornAuron can work with Celeborn as a shuffle manager. Integration involves configuring Auron/Spark to use the AuronCelebornShuffleManager and pointing it to the appropriate Celeborn master endpoints and storage locations. This allows Spark jobs running on Auron to leverage Celeborn for distributed shuffling. You can integrate using the following example configuration: # change celeborn endpoint and storage directory to the correct location
spark.shuffle.manager org.apache.spark.sql.execution.auron.shuffle.celeborn.AuronCelebornShuffleManager
spark.serializer org.apache.spark.serializer.KryoSerializer
spark.celeborn.master.endpoints localhost:9097
spark.celeborn.client.spark.shuffle.writer hash
spark.celeborn.client.push.replicate.enabled false
spark.celeborn.storage.availableTypes HDFS
spark.celeborn.storage.hdfs.dir hdfs:///home/celeborn
spark.sql.adaptive.localShuffleReader.enabled false Apache UniffleSimilarly, Auron supports Uniffle, a high-performance remote shuffle service. To enable Uniffle, you need to configure Auron/Spark to use the AuronUniffleShuffleManager and specify the Uniffle coordinator endpoints. You can integrate using the following example configuration: spark.shuffle.manager org.apache.spark.sql.execution.auron.shuffle.uniffle.AuronUniffleShuffleManager
spark.serializer org.apache.spark.serializer.KryoSerializer
spark.rss.coordinator.quorum <coordinatorIp1>:19999,<coordinatorIp2>:19999
spark.rss.enabled true Notes
|
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
The detailed solution is in the post. |
Beta Was this translation helpful? Give feedback.
The detailed solution is in the post.