[ 
https://issues.apache.org/jira/browse/HADOOP-2000?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Nigel Daley updated HADOOP-2000:
--------------------------------

    Status: Open  (was: Patch Available)

Great!  Almost there. 

* 30 second barrier default I think is too low.  I'd suggest at 1 or 2 minutes.
* remove debug output
* optional: validateInputs() should check for negative parameter inputs that 
don't make sense
* optional: parseInputs() should output startTime in a format that is the 
default used in NN and DN logs
* in barrier(), "if (currentTime <= startTime)" would be clearer as "if 
(sleepTime > 0)"
* totalTimeTPS and startTimeTPS are really TPmS
* f: and s: are never used so that code can be removed from the reducer
* in map(), the openRead, rename, and delete cases should not call 
doCreateWriteOp as this screws up the point of the barrier;  the assumption is 
that the createWrite operation is done before the other operations -- perhaps 
this needs to be in the usage message and javadoc
* rename maxFailedExceptions to MAX_OPERATION_EXCEPTIONS
* shouldn't replication factor be a short throughout?
* shouldn't openRead optionally read the file and record the time for reading?
* I think number of files need to be sent to the reducer since it may be less 
than the desired if the map is late or there are too many exceptioins; this 
should correct the value that is currently used in the analyzeResults() method
* remove tps1and2 and just multiply tps1 (which you then may want to rename) in 
the "If (operation.equals(OP_CREATE_WRITE))"
* document the forumulas in the analyzeResult() method

> Re-write NNBench to use MapReduce
> ---------------------------------
>
>                 Key: HADOOP-2000
>                 URL: https://issues.apache.org/jira/browse/HADOOP-2000
>             Project: Hadoop
>          Issue Type: Test
>          Components: test
>    Affects Versions: 0.15.0
>            Reporter: Mukund Madhugiri
>            Assignee: Mukund Madhugiri
>             Fix For: 0.16.0
>
>         Attachments: HADOOP-2000.patch, HADOOP-2000.patch, HADOOP-2000.patch
>
>
> The proposal is to re-write the NNBench benchmark/test to measure Namenode 
> operations using MapReduce. Two buckets of measurements will be done:
> 1. Transactions per second 
> 2. Average latency
> for these operations
> - Create and Close file
> - Open file
> - Rename file
> - Delete file

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

Reply via email to