[
https://issues.apache.org/jira/browse/HADOOP-2736?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12564152#action_12564152
]
Doug Cutting commented on HADOOP-2736:
--------------------------------------
TestSetFile tests random access. The unit test creates a SetFile containing
10,000 RandomDatum instances, then seeks to sqrt(10,000)=100 random entries in
that file and checks that the expected entry is found.
We could also run it on a MiniDFS cluster, specifying a small block size
(~32kb?) to test random access on HDFS.
We might also specify block compression for a run, to better test seeks within
a block-compressed SequenceFile.
> Random read benchmark for DFS
> -----------------------------
>
> Key: HADOOP-2736
> URL: https://issues.apache.org/jira/browse/HADOOP-2736
> Project: Hadoop Core
> Issue Type: New Feature
> Reporter: Raghu Angadi
> Fix For: 0.17.0
>
>
> We should have at least one random read benchmark that can be run with rest
> of Hadoop benchmarks regularly.
> Please provide benchmark ideas or requirements.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
