why impala memory layout is row base

[许益铭]

hi, everyone!
I find impala row batch is row base(Each row is contiguous memory),  but
spark presto clickhouse is column base(every column is contiguous
memory), Is there any benefit to doing this?