You are viewing a plain text version of this content. The canonical link for it is here.

Posted to user@hadoop.apache.org by ravi teja <ra...@gmail.com> on 2016/06/29 11:52:13 UTC

Namenode Fairscheduler's weight configs

Hi Community,

Have some doubts in the fairscheduler, which was introduced as part of the
QoS (HADOOP-9640) .

The default queue weights for 4 priority queues are 8,4,2,1.

I couldn't find any information about the rational behind choosing these
configs.
Why 8,4,2,1 , why not 64,32,16,8 or 128,64,32,16.

The problem I see with lesser values like (8,4,2,1) , there is queue
switching overhead is very frequent.

And for a client handler size of 400, shouldn't 64,32,16,8 values be good
considering 4 priority levels are being set?


Thanks in advance,
Ravi