Are there some configuration parameters controlling the throughput of HDFS storage location?

Did you measure the storage locations throughput with the MEASURE_LOCATION_PERFORMANCE function?

See:
https://my.vertica.com/docs/9.0.x/HTML/index.htm#Authoring/SQLReferenceManual/Functions/VerticaFunctions/MEASURE_LOCATION_PERFORMANCE.htm

There are a bunch of database parameters for Hadoop:

select parameter_name, description from vs_configuration_parameters where parameter_name ilike '%hdfs%' or parameter_name ilike '%hadoop%';

Do you have Vertica installed on the same cluster as Hadoop? If not, maybe you have a network bottleneck between Vertica and Hadoop?

The parameter HadoopFSReplication controls the number of copies of data HDFS stores. Maybe reducing it to 1 will help you?

See:
https://my.vertica.com/docs/9.0.x/HTML/index.htm#Authoring/HadoopIntegrationGuide/StorageLocations/TroubleSHootingHDFSStorageLocations.htm

Thank you Jim. But it seems HadoopFSReplication came from webhdfs filesystem era, does it work for our NativeHadoopFileSystem?

I assume the example in the doc here:

https://my.vertica.com/docs/9.0.x/HTML/index.htm#Authoring/HadoopIntegrationGuide/StorageLocations/TroubleSHootingHDFSStorageLocations.htm

Is using the a storage location defined using the hdfs:/// URL.

I just sent an email asking the doc team to add a CREATE LOCATION to the example.

=> CREATE LOCATION 'hdfs:///user/dbadmin' ALL NODES SHARED USAGE 'data' LABEL 'hdfs';

I think that would be more clear as the earlier examples create a storage location labeled "coldstorage".