hive-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Dong Chen" <dong1.c...@intel.com>
Subject Re: Review Request 28964: HIVE-8121 Create micro-benchmarks for ParquetSerde and evaluate performance
Date Thu, 15 Jan 2015 07:47:51 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/28964/#review68215
-----------------------------------------------------------


When analyzing the benchmark data, I found it might be good to make each benchmark independent
or finer granularity. Please see the comments below.

How does that sound?


itests/hive-jmh/src/main/java/org/apache/hive/benchmark/storage/ColumnarStorageBench.java
<https://reviews.apache.org/r/28964/#comment112364>

    OrcRecordWriter load its internal writer lazily, which means writing the first record
is different with writing following records.
    
    Should we consider this when defining and evaluating benchmark?



itests/hive-jmh/src/main/java/org/apache/hive/benchmark/storage/ColumnarStorageBench.java
<https://reviews.apache.org/r/28964/#comment112360>

    The benchmark value of this method is sum of: getRecordWriter and writer.write.
    The first one is covered in another benchmark method.
    
    Shall we pass the writer in this method and let it only focus on the second one?



itests/hive-jmh/src/main/java/org/apache/hive/benchmark/storage/ColumnarStorageBench.java
<https://reviews.apache.org/r/28964/#comment112361>

    Same as above comments. Can we change the read method to only test reader.next(), and
remove the getRecordReader part?


- Dong Chen


On Jan. 14, 2015, 11:48 p.m., Sergio Pena wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/28964/
> -----------------------------------------------------------
> 
> (Updated Jan. 14, 2015, 11:48 p.m.)
> 
> 
> Review request for hive, Brock Noland and cheng xu.
> 
> 
> Bugs: HIVE-8121
>     https://issues.apache.org/jira/browse/HIVE-8121
> 
> 
> Repository: hive-git
> 
> 
> Description
> -------
> 
> This is a new tool used to test ORC & PARQUET file format performance.
> 
> 
> Diffs
> -----
> 
>   itests/hive-jmh/pom.xml PRE-CREATION 
>   itests/hive-jmh/src/main/java/org/apache/hive/benchmark/storage/ColumnarStorageBench.java
PRE-CREATION 
>   itests/pom.xml 40e7b9b8114d693f352f28ad94e6abcf96d1372a 
> 
> Diff: https://reviews.apache.org/r/28964/diff/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Sergio Pena
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message