Hi,
I am building Elasticsearch on Microsoft Azure Virtual machines. With Azure
VM, I could build Load Balance on top of all my VMs so that I can wrap all
VMs into one single endpoint exposed for query.
Now my question is, is such external load balance really useful for query
or index performance? Since for each query, any node who receive the
request will forward the query request to all shards which loaded on other
VMs, so seems I can simply always send request to one node VM.
It'd help if you had client nodes.
But it also helps if you don't, even though ES does connect to the nodes
with the shards it needs, distributing the initial query load across
multiple nodes means that you don't have a single node that has to deal
with all the indexing and querying results from the end users/clients.
Hi,
I am building Elasticsearch on Microsoft Azure Virtual machines. With
Azure VM, I could build Load Balance on top of all my VMs so that I can
wrap all VMs into one single endpoint exposed for query.
Now my question is, is such external load balance really useful for query
or index performance? Since for each query, any node who receive the
request will forward the query request to all shards which loaded on other
VMs, so seems I can simply always send request to one node VM.
thanks your reply.
what do you mean by "client nodes"? Do you mean the "NO master and No data”
nodes just work as load balancer?
On Wednesday, April 29, 2015 at 3:27:42 PM UTC+8, Mark Walkom wrote:
It'd help if you had client nodes.
But it also helps if you don't, even though ES does connect to the nodes
with the shards it needs, distributing the initial query load across
multiple nodes means that you don't have a single node that has to deal
with all the indexing and querying results from the end users/clients.
On 29 April 2015 at 17:12, Xudong You <xudon...@gmail.com <javascript:>>
wrote:
Hi,
I am building Elasticsearch on Microsoft Azure Virtual machines. With
Azure VM, I could build Load Balance on top of all my VMs so that I can
wrap all VMs into one single endpoint exposed for query.
Now my question is, is such external load balance really useful for query
or index performance? Since for each query, any node who receive the
request will forward the query request to all shards which loaded on other
VMs, so seems I can simply always send request to one node VM.
thanks your reply.
what do you mean by "client nodes"? Do you mean the "NO master and No
data” nodes just work as load balancer?
On Wednesday, April 29, 2015 at 3:27:42 PM UTC+8, Mark Walkom wrote:
It'd help if you had client nodes.
But it also helps if you don't, even though ES does connect to the nodes
with the shards it needs, distributing the initial query load across
multiple nodes means that you don't have a single node that has to deal
with all the indexing and querying results from the end users/clients.
Hi,
I am building Elasticsearch on Microsoft Azure Virtual machines. With
Azure VM, I could build Load Balance on top of all my VMs so that I can
wrap all VMs into one single endpoint exposed for query.
Now my question is, is such external load balance really useful for
query or index performance? Since for each query, any node who receive the
request will forward the query request to all shards which loaded on other
VMs, so seems I can simply always send request to one node VM.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.