Hi All,
I have seen some intrest in clustering but I see that there is no ongoing discussion or any clear requirments on this topic.
What I would like to do is to start some discussion and gather some momentum around this so we can flush out the requirments/expectations of clustering for Axis2.
I can volunteer to cordinate the effort, but I really need people to come out with suggestions and expectations.
To start with I would like to pose a few questions and get the ball rolling...
1. What is the basic unit of clustering from Axis2 point of view ?
2. Since web services are more xml driven than object driven, can a object focused clustering solution be used within Axis2 ?
3. What are some of the challengers or limitations we already know ?
4. If we consider a homogenous collection of web services deployed on several nodes what will be the users expectations on clustering?
a.) Configuration data to be clustered to all nodes vs One time manul update ???
b.) If a web service fails during a request how intelligent should axis2 be in determining wether to transparently move the session/request to another node?
c.) What policies are we going to support ??? Sticky sessions vs no-sticky sessions and several other variations of this.
We should provide an API for people to write there own policy/impl for clustering stratergy.
d.) we support several transports, therefore how can we generalize our clustering support to be applicable within all transports ??
Comments/suggestions are very welcomed !!!
Regards,
Rajith Attapattu.
