Our Community is getting an upgrade! To get everything ready for the relaunch, we’ll be placing the site in read-only mode starting September 21st.
We really appreciate your understanding while we get things set up behind the scenes. Catch up on all the exciting details about the move here.
Need help or have questions? Drop us a line at [email protected]
Created 01-03-2018 02:36 PM
Hi,
I know that Tez avoids storing intermediates result into HDFS (versus MapReduce that does it) but I was wondering, where are they stored then ?
I read : "on memory", "on local disk"...
But what if the task which emits intermediates result are not on the same node that the task which will receive it ?
So, is it just network I/O instead of HDFS read / write streaming datas from memory and/or local disk ?
Thanks for your help 🙂
Created 01-04-2018 09:27 PM
There is some information here about data movement: https://hortonworks.com/blog/writing-a-tez-inputprocessoroutput-2/
Tez is pluggable and has different data transfer paradigms, but in general things are kept in memory until size constraints cause flushes to local disk. When tasks are not on the same node data will be transferred over the network (out of band data movement events involving the AM + direct data transfer between the nodes.)
Created 01-04-2018 09:27 PM
There is some information here about data movement: https://hortonworks.com/blog/writing-a-tez-inputprocessoroutput-2/
Tez is pluggable and has different data transfer paradigms, but in general things are kept in memory until size constraints cause flushes to local disk. When tasks are not on the same node data will be transferred over the network (out of band data movement events involving the AM + direct data transfer between the nodes.)
Created 01-06-2018 08:39 AM
Hi @Gunther Hagleitner ; thanks it's very clear with your explainations.