[
https://issues.apache.org/jira/browse/HIVE-3746?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13648667#comment-13648667
]
Carl Steinbach commented on HIVE-3746:
--------------------------------------
bq. If an application has requested a single row, and the client has requested
n rows from the server in an effort to reduce round trips, then n-1 intervening
values from the first column must be cached off somewhere before the first
value for the second column can be accessed.
If the fetch size is n, then the client is going to end up storing n rows in
memory regardless of whether the result set is represented in a row-major or
column-major format. Put another way, the unit of data transfer between the
server and client is a variable sized resultset. The client has the option of
setting the result size very low in order to achieve lower latency, or making
it very large in order to get higher overall throughput. However, the key
limitation is that the client is not able to provide access to any of the rows
contained in a resultset until the entire resultset has been transferred from
the server to the client. This limitation is a consequence of the fact that
we're using a message oriented RPC layer (Thrift) to handle communication and
data transfer between the client and server.
> TRowSet resultset structure should be column-oriented
> -----------------------------------------------------
>
> Key: HIVE-3746
> URL: https://issues.apache.org/jira/browse/HIVE-3746
> Project: Hive
> Issue Type: Sub-task
> Components: Server Infrastructure
> Reporter: Carl Steinbach
> Assignee: Carl Steinbach
> Labels: HiveServer2
>
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
