Do we need a Nodebuilder pool


(CHI WAI HO) #1

Hi All,

I am new to EC and have just gone through some tests it it. It is
doing really great on it's scability and performance. However, as I am
originally a RDBMS guy who used JDBC And sone pooling mechanism to
reduce the connection establishment time, is there any equivalent
things that are available with EC? Or the nodebuilder will do the
pooling (I am mentioning about the pooling, not the dynamic node
discovery) for us? As making new connection for each user request is
expensive and lower the response rate I guess.

May be there are topics already mentioned the mechanism on EC website
that I missed, if there is any, would you mind sharing the links with
us?

Thank you very much and you every kind help and reply are highly
appreciated .

Zero


(medcl.net) #2

pooling is needed,when you are using thrift,it's easy to reuse the socket
connection,and the performance is burst up,I've already implemented that in
my ElasticSearch.NET Client,you can check it out:
https://github.com/medcl/ElasticSearch.Net

-----Original Message-----
From: CHI WAI HO
Sent: Thursday, August 11, 2011 11:01 AM
To: elasticsearch
Subject: Do we need a Nodebuilder pool

Hi All,

I am new to EC and have just gone through some tests it it. It is
doing really great on it's scability and performance. However, as I am
originally a RDBMS guy who used JDBC And sone pooling mechanism to
reduce the connection establishment time, is there any equivalent
things that are available with EC? Or the nodebuilder will do the
pooling (I am mentioning about the pooling, not the dynamic node
discovery) for us? As making new connection for each user request is
expensive and lower the response rate I guess.

May be there are topics already mentioned the mechanism on EC website
that I missed, if there is any, would you mind sharing the links with
us?

Thank you very much and you every kind help and reply are highly
appreciated .

Zero


(Shay Banon) #3

There is no need to pool Node instances (NodeBuilder is just a way to
construct a node instances), you should have one Node instances across the
lifecycle of the app used by as many threads as you want.

On Thu, Aug 11, 2011 at 6:01 AM, CHI WAI HO zeroho@gmail.com wrote:

Hi All,

I am new to EC and have just gone through some tests it it. It is
doing really great on it's scability and performance. However, as I am
originally a RDBMS guy who used JDBC And sone pooling mechanism to
reduce the connection establishment time, is there any equivalent
things that are available with EC? Or the nodebuilder will do the
pooling (I am mentioning about the pooling, not the dynamic node
discovery) for us? As making new connection for each user request is
expensive and lower the response rate I guess.

May be there are topics already mentioned the mechanism on EC website
that I missed, if there is any, would you mind sharing the links with
us?

Thank you very much and you every kind help and reply are highly
appreciated .

Zero


(system) #4