hadoop-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Devaraj k <devara...@huawei.com>
Subject RE: Map slots and Reduce slots
Date Mon, 15 Jul 2013 02:10:08 GMT
These configurations cannot be changed dynamically. We need to configure these values for Tasks
Tracker's before starting and cannot be changed after that. If we want to change these then
TT's need to be restarted. You can configure the cluster based on the resources available.
You can tune your Job configuration according your cluster configurations.

Thanks
Devaraj k

From: Shekhar Sharma [mailto:shekhar2581@gmail.com]
Sent: 15 July 2013 07:32
To: user@hadoop.apache.org
Subject: Re: Map slots and Reduce slots

Sorry for the wrong properties name, i meant the same..
I understand the properties functionality, can i add the slots at run time to a particular
task trackers depending on the load, because as you suggested we can determine the slots depending
on the load..and since the load can be dynamic, so can i dynamically allocate the task tracker
based on information lets say depending on the availablity of the resources on the task tracker
machine..

Regards,
Som Shekhar Sharma
+91-8197243810

On Mon, Jul 15, 2013 at 7:27 AM, Devaraj k <devaraj.k@huawei.com<mailto:devaraj.k@huawei.com>>
wrote:
Hi Shekar,

   I assume you are trying with Hadoop-1. There are no properties with the names 'mapred.map.max.tasks'
and 'mapred.reduce.max.tasks'.

We have these configuration to control the max no of map/reduce tasks run simultaneously.
mapred.tasktracker.map.tasks.maximum - The maximum number of map tasks that will be run simultaneously
by a task tracker.
mapred.tasktracker.reduce.tasks.maximum - The maximum number of reduce tasks that will be
run simultaneously by a task tracker.

For ex: If we declare mapred.tasktracker.map.tasks.maximum=3  and mapred.tasktracker.reduce.tasks.maximum=4
for a task tracker, means the TT has 3 map slots and 4 reduce slots.

> Let's say on a machine if i have 8GB RAM and dual core machine, how can i determine that
what would be the optimal number of map and reducer slots for this machine
It purely depends on which type of tasks you are going to run and load of the task. Normally
each task requires one core to execute, no of concurrent tasks can be configured based on
this. And memory required for the task depends on how much data it is going to process.


Thanks
Devaraj k

From: Shekhar Sharma [mailto:shekhar2581@gmail.com<mailto:shekhar2581@gmail.com>]
Sent: 14 July 2013 23:15
To: user@hadoop.apache.org<mailto:user@hadoop.apache.org>
Subject: Map slots and Reduce slots

Does the properties mapred.map.max.tasks=3 and mapred.reduce.max.tasks=4 means that machine
has 3 map slots and 4 reduce slots?

Or is there any way i can determine the number of map slots and reduce slots that i can allocate
for a machine?

Let's say on a machine if i have 8GB RAM and dual core machine, how can i determine that what
would be the optimal number of map and reducer slots for this machine




Regards,
Som Shekhar Sharma
+91-8197243810<tel:%2B91-8197243810>


Mime
View raw message