Is there a way to adjust receiving buffer of COPY between nodes?

Hi team,

I have a prospect who is evaluating Vertica. But he met a performance issue when COPY data. There are low CPU/IO/network usages when 3 COPYs are loading data from local files, and no any bottleneck. No abnormal result from vioperf/vnetperf checking. All network bandwidth can be used when COPYs going to phase II. His statement looks like:

    COPY ... from
    'file1' on node1,
    'file2' on node2,
    'file3' on node3,
    -- 'file4' on node4,
    -- 'file5' on node5,
    -- 'file6' on node6,
    -- 'file7' on node7,
    'file8' on node8,
    'file9' on node9,
    'file10' on node10
    ... ;

The weird thing is, when he comments out about any 4~6 files/nodes in his COPY statements, IO/network usages can achive good higher numbers! It looks like there is some bottleneck when receiving data from other nodes.

Do you know whether there is a way to adjust receiving buffer of COPY between nodes?

Regards,
DQ