HDFS ingest rate

[steph]

Hi,

Are there any performance numbers related to how HDFS can ingest data?
I am assuming a case where multiple processes outside hadoop write into
hadoop in parallel. I understand that it is probably related to  
various hardware
constraints but any existing numbers would be interesting. In  
particular:
Is the ingest rate tied to the number of nodes on the cluster? Is it  
somehow linear?
What is the impact of ingesting data at a high rate on the map/reduce  
jobs running?

Thanks,

S.