Load balancer in elasticsearch

Hello everyone,
I am looking to restructure my ES architecture. I would like to have something like:

MACHINE 1 (Coordinator) --> MACHINE 2
--> MACHINE 3

A virtual machine (Machine 1) that its mission is to redirect the data to 2 servers (MACHINE 2 , MACHINE 3) and this machine 1 does not have data.
These two servers have a lot of memory a hard disk. so, I would like to separate the information in two servers and ask to the coordinator machine (MACHINE 1) for a query and the coordinator will ask to the 2 servers and return the answer/hits.
In other words, If I have 5000 docs, I would have 2500 documents in MACHINE 2 and 2500 documents in MACHINE 3 and if I ask to MACHINE 1 for a query it will return the answer merged between MACHINE 2 and MACHINE 3.
Is it possible? or is it another way two separate the information in two different servers?

Thankssssss

A client node can do that - https://www.elastic.co/guide/en/elasticsearch/reference/2.4/modules-node.html#client-node