[
https://issues.apache.org/jira/browse/MAPREDUCE-961?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12777644#action_12777644
]
Scott Chen commented on MAPREDUCE-961:
--------------------------------------
Thanks for the comment, Arun. I have changed the patch a lot following the
suggestion froms Matei and Vinod. The last patch is total different from the
first one. I am sorry about the confusion.
The following is the design
1. We obtain the available memory on the TT using MemoryCalculatorPlugin.
Originally this class calculates only total memory only, we add a slight change
so that it also computes the available memory.
2. The information is reported with TaskTrackerStatus.ResourceStatus back to JT.
3. In MemBasedLoadManager, we look at the available memory on TT, the maximum
memory per task (from jobConf) and a configured reserved memory on TT. If
(available memory - task memory > reserved memory), we return true which allows
scheduler to lauch the task.
The initial idea also includes using the memory usage of a job collecting in
the cluster. Right now we only use the value obtained in jobConf. After
MAPREDUCE-220 is done, we can use the task memory estimated by the previous
tasks.
> ResourceAwareLoadManager to dynamically decide new tasks based on current
> CPU/memory load on TaskTracker(s)
> -----------------------------------------------------------------------------------------------------------
>
> Key: MAPREDUCE-961
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-961
> Project: Hadoop Map/Reduce
> Issue Type: New Feature
> Components: contrib/fair-share
> Affects Versions: 0.22.0
> Reporter: dhruba borthakur
> Assignee: Scott Chen
> Fix For: 0.22.0
>
> Attachments: HIVE-961.patch, MAPREDUCE-961-v2.patch,
> MAPREDUCE-961-v3.patch, MAPREDUCE-961-v4.patch
>
>
> Design and develop a ResouceAwareLoadManager for the FairShare scheduler that
> dynamically decides how many maps/reduces to run on a particular machine
> based on the CPU/Memory/diskIO/network usage in that machine. The amount of
> resources currently used on each task tracker is being fed into the
> ResourceAwareLoadManager in real-time via an entity that is external to
> Hadoop.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
