See the JMS Performance Guide white-paper on dev2dev. Pay
special attention to the pooling and clustering sections.

Radhakrishnan J wrote:

> Hi,
>
> I would like to know what is the best approach for managing the JMS connections
> when the client is on a remote machine ? I've tried declaring the JMS connection
> factory as a Servlet or EJB resource for the following scenarios
>
> (1) The application is deployed on a cluster with a connection factory deployed
> on the cluster.
> (2) The application is deployed on a cluster, but the connection factory is deployed
> on only 1 node of the cluster. In this case, it doesn't matter from which node
> the connections are being obtained.
>
> In our case, we'd have Weblogic Express instances using JMS service hosted on
> WLS. Should there be a connection connection pooling in this case ?
>
> Thanks,
> Radhakrishnan
>
>

Tom,

Thanks for the pointer. I'm yet to go through the complete document, herez a quick
question:

From what I gather, remote clients obtaining connections to WebLogic server(s),
that are clustered, should not pool those connections.

I have a more basic question. I am more eager to know what happens at the basic
level. Let me run through the steps that a remote client would do to obtain a
JMS connection

- Create a configured initial context for JNDI lookups.
- By the time a lookup for the connection factory is served, a TCP connection
is established with a WL server for the JNDI service. If there is a cluster in
place, There would be a load-balancing ( either at the DNS level or if the URL
is a comma separated list of member servers) that would determine to which server
the TCP connection gets established.
- When the JMS connection is created using the connection factory
(a) Is this a new TCP connection ? If so to which server ?
OR
(b) Is the same TCP connection setup while doing JNDI lookup used for the JMS
connection as well ?

Thanks,
Radhakrishnan

Tom Barnes
wrote:
>See the JMS Performance Guide white-paper on dev2dev. Pay
>special attention to the pooling and clustering sections.
>
>Radhakrishnan J wrote:
>
>> Hi,
>>
>> I would like to know what is the best approach for managing the JMS
>connections
>> when the client is on a remote machine ? I've tried declaring the JMS
>connection
>> factory as a Servlet or EJB resource for the following scenarios
>>
>> (1) The application is deployed on a cluster with a connection factory
>deployed
>> on the cluster.
>> (2) The application is deployed on a cluster, but the connection factory
>is deployed
>> on only 1 node of the cluster. In this case, it doesn't matter from
>which node
>> the connections are being obtained.
>>
>> In our case, we'd have Weblogic Express instances using JMS service
>hosted on
>> WLS. Should there be a connection connection pooling in this case ?
>>
>> Thanks,
>> Radhakrishnan
>>
>>
>

- a TCP/IP connection is established when establishing the context:
term the host WL server the "context host"

- between a client and a particular wl server there is only
one TCP/IP connection, all EJB/JMS/JNDI traffic is
multiplexed over this connection

- the JMS connection factory may return a connection to a WL server
host other
than the context host, which, as a result, creates
a new TCP/IP connection (see the clustering
section of the white-paper for details)

- pooling of JMS connections/sessions/producers/destinations
etc is recommended if high performance is needed. (Hundreds
or thousands established per second.)


Radhakrishnan J wrote:

> Tom,
>
> Thanks for the pointer. I'm yet to go through the complete document, herez a quick
> question:
>
> From what I gather, remote clients obtaining connections to WebLogic server(s),
> that are clustered, should not pool those connections.
>
> I have a more basic question. I am more eager to know what happens at the basic
> level. Let me run through the steps that a remote client would do to obtain a
> JMS connection
>
> - Create a configured initial context for JNDI lookups.
> - By the time a lookup for the connection factory is served, a TCP connection
> is established with a WL server for the JNDI service. If there is a cluster in
> place, There would be a load-balancing ( either at the DNS level or if the URL
> is a comma separated list of member servers) that would determine to which server
> the TCP connection gets established.
> - When the JMS connection is created using the connection factory
> (a) Is this a new TCP connection ? If so to which server ?
> OR
> (b) Is the same TCP connection setup while doing JNDI lookup used for the JMS
> connection as well ?
>
> Thanks,
> Radhakrishnan
>
> Tom Barnes
> wrote:
>
>>See the JMS Performance Guide white-paper on dev2dev. Pay
>>special attention to the pooling and clustering sections.
>>
>>Radhakrishnan J wrote:
>>
>>
>>>Hi,
>>>
>>>I would like to know what is the best approach for managing the JMS
>>
>>connections
>>
>>>when the client is on a remote machine ? I've tried declaring the JMS
>>
>>connection
>>
>>>factory as a Servlet or EJB resource for the following scenarios
>>>
>>>(1) The application is deployed on a cluster with a connection factory
>>
>>deployed
>>
>>>on the cluster.
>>>(2) The application is deployed on a cluster, but the connection factory
>>
>>is deployed
>>
>>>on only 1 node of the cluster. In this case, it doesn't matter from
>>
>>which node
>>
>>>the connections are being obtained.
>>>
>>>In our case, we'd have Weblogic Express instances using JMS service
>>
>>hosted on
>>
>>>WLS. Should there be a connection connection pooling in this case ?
>>>
>>>Thanks,
>>>Radhakrishnan
>>>
>>>
>>
>