While executing scalability tests the DM somethimes writes the following lines to the terminal. It seems like this does not influence the execution of the operation. Unfortunately, I cannot provide further information because logging is completely disabled while executing the benchmarks.
Exception in fileStream Message :IMcsv_GWuriy_1.csv of size 141481 to host 10.254.142.69:1805(domU-12-31-39-00-65-06.compute-1.internal:1805)
Exception in fileStream Message :IMcsv_GWuriy_1.csv of size 141481 to host 10.254.203.192:1805(domU-12-31-39-00-65-06.compute-1.internal:1805)
Exception in fileStream Message :IMcsv_ByYVrP_2.csv of size 142010 to host 10.254.203.192:1805(domU-12-31-39-00-65-06.compute-1.internal:1805)
Exception in fileStream Message :IMcsv_ByYVrP_2.csv of size 142010 to host 10.254.142.69:1805(domU-12-31-39-00-65-06.compute-1.internal:1805)
Exception in fileStream Message :IMcsv_HoDURK_3.csv of size 141896 to host 10.254.203.192:1805(domU-12-31-39-00-65-06.compute-1.internal:1805)
Exception in fileStream Message :IMcsv_HoDURK_3.csv of size 141896 to host 10.254.142.69:1805(domU-12-31-39-00-65-06.compute-1.internal:1805)
This is related to P-Grid implementation. As you also mentioned it does not influence the execution of read or out operation. It is only because of too many loads on one single node.
the out operation stops working while measuring scalability on ec2 (it seems like it does not return). The exception shown in the initial comment is the only exception which is thrown.
Is it possible that this behaviour is related to the exception?