You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@turbine.apache.org by Gustavo De Simone <gu...@swordfish.com.uy> on 2001/04/30 20:39:25 UTC
I need your help
Hi, my name is Gustavo De Simone I am taking my thesis with Andrea Antelo at
the University Facultad de Ingenieria of Uruguay on our final year. We have
to evaluate Turbine-Tomcat as a tool for building Applications and running
them so we have some questions to make you.
We have recently develop a small application that use tomcat,turbine, torque
and velocity .Resuming our experience with it, we found it really
usefull,easy to use and fast to develope.
But, due to we don't have the necesary equipment we couldn't make any
measurement about runtime issues.So I would like to make to all of you some
questions:
Scalability:
We suppose an scenario were the quantity of request has grown in a really
big number ,and where we are expecting that it grows more. In that case we
need our tool to update to this needs.
Our proposal about this is the next :
We point to have something like this
watch Figure 1
Actually sessions are persistents between requests, but that persistents is
stored in JVM memory, so if a request comes for the first time to machine1
and consecuently a session is created, that information is not share by the
others machines so others request from that user could only be responsed by
the machine1.
A possible solution may be to store sessions and the associated object to
database, so sessions doesn 't depend on the Virtual Machine. We know that
Turbine get into the session a lot of security issue for example ACL, but
all of that may be made persistent.
Another way (Figure 3) could be to have a service that share sessions and
all related objects making rmi calls. In this case before create a session
we should ask to other computers if there is not that session on their
engines.Of course in both cases the load balancer doesn't matter about
session it only sends request to the WebServer least charged.
We tried to make this e-mail as short as we could so if you think that it is
not enough explained please let us know.
We suppose that you have already thought on this same solutions so we like
to know your opinion.
It would be really important to our thesis to know what you think and if you
think this could be a good solution we'd like to work on that.
with regards
Gustavo y Andrea