Mailing-List: contact hadoop-dev-help@lucene.apache.org; run by ezmlm
Precedence: bulk
Reply-To: hadoop-dev@lucene.apache.org
Message-ID: <28846478.1198004863014.JavaMail.jira@brutus>
Date: Tue, 18 Dec 2007 11:07:43 -0800 (PST)
From: "dhruba borthakur (JIRA)" <jira@apache.org>
To: hadoop-dev@lucene.apache.org
Subject: [jira] Updated: (HADOOP-2434) MapFile.get on HDFS in TRUNK is
 WAY!!! slower than 0.15.x
In-Reply-To: <28165661.1197685303191.JavaMail.jira@brutus>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


     [ https://issues.apache.org/jira/browse/HADOOP-2434?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

dhruba borthakur updated HADOOP-2434:
-------------------------------------

    Resolution: Fixed
        Status: Resolved  (was: Patch Available)

I just committed this. Thanks Raghu!

> MapFile.get on HDFS in TRUNK is WAY!!! slower than 0.15.x
> ---------------------------------------------------------
>
>                 Key: HADOOP-2434
>                 URL: https://issues.apache.org/jira/browse/HADOOP-2434
>             Project: Hadoop
>          Issue Type: Bug
>          Components: io
>            Reporter: stack
>            Assignee: stack
>            Priority: Blocker
>             Fix For: 0.16.0
>
>         Attachments: 2434.patch, TestMapFile.java, TestMapFile.java
>
>
> Stall happens down in SequenceFile in the first call to getPos inside readRecordLength.  I tried the johano patch from HADOOP-2172 that restores the positional cache but that didn't seem to be the issue here.
> Here is data to support my assertion.
> I wrote a little program to make a MapFile of 1M records.  I then did 1M random reads from same file.  Below are timings from a 0.15.0 and TRUNK as of this afternoon run.
> 0.15.x branch:
> {code}
> [stack@aa0-000-12 branch-0.15]$ ./bin/hadoop org.apache.hadoop.io.TestMapFile
> .07/12/15 01:29:02 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
> 07/12/15 01:29:02 INFO io.TestMapFile: Writing 1000000 rows to testMapFileRandomAccess
> 07/12/15 01:32:04 INFO io.TestMapFile: Writing 1000000 records took 182009ms
> 07/12/15 01:32:04 INFO io.TestMapFile: Reading 1000000 random rows
> 07/12/15 01:48:02 INFO io.TestMapFile: Reading 1000000 random records took 958243ms
> Time: 1,140.652
> OK (1 test)
> {code}
> For the below test using TRUNK r604352, I amended the test so it output a log message every 100k reads:
> {code}
> [stack@aa0-000-12 hadoop-trunk]$ ./bin/hadoop org.apache.hadoop.io.TestMapFile
> .07/12/15 01:56:34 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
> 07/12/15 01:56:34 INFO io.TestMapFile: Writing 1000000 rows to testMapFileRandomAccess
> 07/12/15 01:59:38 INFO io.TestMapFile: Writing 1000000 records took 183986ms
> 07/12/15 01:59:38 INFO io.TestMapFile: Reading 1000000 random rows
> .........
> {code}
> After 20 minutes it still hadn't printed out the 'read 100k messages' (I had to leave -- will fill in final figures later)

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.