tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Pid <...@pidster.com>
Subject Re: multi-tented connection queues
Date Tue, 10 Jan 2012 14:25:29 GMT
On 10/01/2012 12:25, Tharindu Patikirikorala wrote:
> Hi All,
> 
> 
> I am doing a case study to make tomcat to host multi-tented web
> applications. i.e., a single web application/service container is serving
> logically partitioned customers.  For the basic implementation (let’s say
> number of customers/tenant are 2) I want to make following modifications.
> 
> 1)      Identify the tenant and implementing a request/connection queues
> for each tenant. Then queue will have ‘acceptcount’ which reject the system
> to avoid severe overloads.

Queue of what?  Tomcat doesn't have any queues.  Queuing is usually
associated with messaging products or applications.


> 2)      The worker threads have to be allocated to serve these requests
> based on a new scheduling. Two queues will be processed by this scheduler
> based on a priority scheme.

Sounds like an application problem.  A messaging application problem.


> After looking at the code, currently a single connection queue
> (ServerSocket.listen(backlog)) is used which is not aware about the tenants
> request. However, for the implementation we need another level of queuing
> which will reject the requests based on the tenant.

You are referring to the Connector.  The Connector processes requests.
You could extend it yourself to do whatever you wanted, but you are
probably just better creating a separate instance for each tenant and
starting it in its own JVM.

See RUNNING.txt for details of how to separate CATALINA_HOME and
CATALINA_BASE so you can use a common runtime for multiple instances.


p


> Please suggest a way to implement such a scheme.
> 
> 
> Thank you
> 
> Pati
> 


-- 

[key:62590808]


Mime
View raw message