Thanks a lot Eric. I will try with HIVE-4483 patch. As you mentioned, It would be awesome to update the standard input formats to leverage vectorization.
~Rajesh.B On Fri, Jan 10, 2014 at 1:23 AM, Eric Hanson (BIG DATA) < eric.n.han...@microsoft.com> wrote: > There’s actually a different inputformat for vectorized processing on > RCFile. See https://issues.apache.org/jira/browse/HIVE-4483. Vectorized > execution won’t run as fast on RCFile as ORC, but there should still be a > noticeable improvement on RCFile. > > > > In the future, I think it’s best to update the standard input formats, so > they can work vectorized or row-at-a-time. This makes for easier evolution > to allow vectorization to run against existing tables. This was done for > ORC. > > > > I’m not sure how deep the testing was on running queries using the > inputformat from HIVE-4483 with RC File. It is much less than for > vectorized query on ORC. > > > > Eric > > > > *From:* Rajesh Balamohan [mailto:rajesh.balamo...@gmail.com] > *Sent:* Wednesday, January 8, 2014 6:47 PM > *To:* user@hive.apache.org > *Subject:* Vectorizied execution on RCFile > > > > Hi All, > > Vectorization with ORCFile provides amazing performance. Does > vectorization work with RCFile as well? > > As per explain plan of Hive 0.13 (snapshot), it does not use vectorization > with RCFile. Any pointers would be appreciated. > > > > > -- > ~Rajesh.B > -- ~Rajesh.B
