Elasticsearch performance question


(Tae Sung Kim) #1

Hi~
I have some question.

First, Which is more effective bulk or indexing?

  1. 1 High spec machine (master + data node)
    ex) 8 core / 32 GB ram / SSD and so on
  2. 3 Normal spec machine (data node) + 1 Low spec machine (master node)
    ex) 2 or 4 core / 4 GB ram / HDD and so on

Second, bulk thread pool is changeable?
Next week, I will test bulk thread pool change.
Now, my machine is 2core 14GB ram and HDD 1TB * 2.
Bulk thread pool min, max is 2.
I want bulk max thread pool to 4.
Is it possible?

If you have experience, I would appreciate your advice.
Thank you.


(chenjinyuan87) #2

For the first question, I've no idea..
For the second, you can change threadpool settings:
https://www.elastic.co/guide/en/elasticsearch/reference/master/modules-threadpool.html


#3

When I using ES1.x whith default configuration, index size is 4x greater then lucene 4.10's index.
It means storage performance is very important in ES.
And smaller memory will calling more JVM GC
In my opinion, 1 case may shows better performance then 2 in your ex specs.


(system) #4

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.